{"approach":"Returned the crossing-cover base code byte-for-byte unchanged as directed; measured paired diff.mean is exactly 0. None of 15 feature tilts passed both the validation-error and ESS criteria on two disjoint 500-run tuning halves, so the frozen proxy retained uniform weights. A separately labelled bounded 18-feature diagnostic fit to all six 4000-run holdout batches is reported with per-file errors, ranking and ESS; it is in-sample calibration, not a validated replacement test.","blocked":null,"python_version":"3.12.3","base_mean_edge":108.63974289615767,"base_unseen":{"mean_edge":97.54703502577897,"retail_edge":126.92890288111944,"arb_edge":-29.381867855334075},"score":{"mean_edge":98.0996956804193,"retail_edge":125.17961744069841,"arb_edge":-27.079921760293374,"simulations":4000,"seed_start":4000,"failures":0},"holdout":{"mean_edge":97.54703502577897,"retail_edge":126.92890288111944,"arb_edge":-29.381867855334075,"simulations":4000,"seed_start":50000,"failures":0},"diff":{"mean":0.0,"standard_error":0.0,"retail":0.0,"arb":0.0},"seconds":{"base":504.6526717925444,"yours":513.9856903129257},"code":"\"\"\"Crossing-funded inventory for passive prediction-market quotes.\n\nUses tiny tick probes, a sub-tick probability filter, and marginal retail-versus-arbitrage sizing to quote one tick inside the competitor. Starting from the exactly reproduced base, adds demand-matched crossing cover at ticks 2-9. When estimated probability is between 2c and 10c and YES inventory is below the desired ordinary ask quantity, a bid rests at floor(100*p), sized to the inventory shortfall, capped by 30% of cash-plus-YES marked wealth and 3000 shares. This gets additional stock on a downward crossing rather than paying the next higher tick, while retaining the base tick-1 reservoir and other quoting rules. Higher fixed-tick reservoirs, broad surplus asks, cash gates, and alternative stock targets were compared on tuning seeds only; selection used the full 600-seed tuning mean. The code was frozen before the holdout.\n\"\"\"\nimport math\nfrom orderbook_pm_challenge.strategy import BaseStrategy\nfrom orderbook_pm_challenge.types import CancelOrder, PlaceOrder, Side\n\nP = dict(lam_r=0.25, m=5.0, sig_n=1.2, lam_j=0.0019, sig_j=0.35, sig=0.02,\n         qmax=3000.0, rmin=2, kA=0.7, kL=0.15, pL=0.05,\n         ammo=1.0, ydem=0.4, W=200.0, fy=0.6, ammo_q=60.0, ammo_pmax=0.05, ammo_maxpay=0.02,\n         fb=0.9, fu=0.7, xu=0.1, sticky=1.0, min_ahead=2.0, keep_hi=1.5, keep_lo=0.7, Hu=300.0, bins=8.0, filt=1.0, uni=0.01, jumpinfo=1.0, dmin=0.0, kLb=0.08, kLf=0.08, pinf=1.0, pre=0.0, H1=100.0,\n         ydf=0.6, Wf=2000.0, g_hi=0.15, g_lo=0.15, qf=400.0, pmaxf=0.05, pre_gap=1.0,\n         ydf1=0.6, g1=5.0, f1=0.6, a1=100.0, nb1=0.0, gs=1.0, p1t=1.0, m1=1.0, zd=1.0, pfl=0.0005)\nSQ2 = math.sqrt(2.0)\nSQ2PI = math.sqrt(2.0 * math.pi)\ndef cdf(x): return 0.5 * (1.0 + math.erf(x / SQ2))\ndef pdf(x): return math.exp(-0.5 * x * x) / SQ2PI\ndef inv(p):\n    p = min(max(p, 1e-9), 1 - 1e-9)\n    t = math.sqrt(-2.0 * math.log(min(p, 1 - p)))\n    x = t - (2.515517 + 0.802853 * t + 0.010328 * t * t) / (1 + 1.432788 * t + 0.189269 * t * t + 0.001308 * t ** 3)\n    x = -x if p < 0.5 else x\n    for _ in range(2):\n        d = pdf(x)\n        if d < 1e-12:\n            break\n        x -= (cdf(x) - p) / d\n    return x\ndef lpm(k):\n    return pdf(k) + k * cdf(k)\ndef fl(q):\n    return math.floor(q * 100 + 1e-9) / 100\n\nNODES = [(-2.2, .06), (-1.3, .17), (-0.6, .27), (0.6, .27), (1.3, .17), (2.2, .06)]\nPROBE = 0.01\n\n\nclass Strategy(BaseStrategy):\n    def __init__(self):\n        self.n = 0\n        self.ahead = set()     # main orders placed strictly ahead of any competitor order at their tick\n        self.res = {}          # order id -> uncovered quantity at placement\n        self.bc = None         # belief: sub-tick grid centres (ticks)\n        self.bw = None         # belief: weights\n        self.st = 0.5          # last one-step sd in ticks\n        self.pcb = self.pca = None\n        self.pb = set()\n        self.pa = set()\n        self.tb = 0\n        self.s = None          # competitor spread parameter\n\n    def _oid(self, tag):\n        self.n += 1\n        return tag + str(self.n)\n\n    def on_step(self, state):\n        H = max(1, state.steps_remaining)\n        cb = state.competitor_best_bid_ticks or 0\n        ca = state.competitor_best_ask_ticks or 100\n        if self.s is None:\n            self.s = max(1, (ca - cb + 1) // 2)\n        lo, hi = cb, ca\n        for o in state.own_orders:\n            if o.side is Side.BUY:\n                lo = max(lo, o.price_ticks)\n            else:\n                hi = min(hi, o.price_ticks)\n        if P['pinf'] > 0:\n            sb = {o.price_ticks for o in state.own_orders if o.side is Side.BUY and o.remaining_quantity <= PROBE + 1e-9}\n            sa = {o.price_ticks for o in state.own_orders if o.side is Side.SELL and o.remaining_quantity <= PROBE + 1e-9}\n            ea = self.pa - sa\n            eb = self.pb - sb\n            if sa and ea:\n                lo = max(lo, min(max(ea), hi))\n            if eb and (sb or min(eb) <= P['p1t']):\n                hi = min(hi, max(min(eb), lo))\n        if P['jumpinfo'] > 0 and self.pcb is not None:\n            if ca > self.pca + 1:\n                lo = max(lo, ca - 1)\n            if cb < self.pcb - 1:\n                hi = min(hi, cb + 1)\n        self.pcb, self.pca = cb, ca\n        if lo > hi:\n            lo, hi = hi, lo\n        # sub-tick belief over p_(t-1): predict with one-step noise, truncate to [lo, hi]\n        wid = max(hi - lo, 1e-6)\n        nb = max(2, min(48, int(round(wid * P['bins']))))\n        cs = [lo + (i + 0.5) * wid / nb for i in range(nb)]\n        if self.bc is None or P['filt'] <= 0:\n            ws = [1.0 / nb] * nb\n        else:\n            sg = max(self.st, 0.6 * wid / nb)\n            k = -0.5 / (sg * sg)\n            ws = []\n            uni = P['uni'] * sg\n            for c in cs:\n                t = uni\n                for ci, wi in zip(self.bc, self.bw):\n                    d = c - ci\n                    d2 = d * d * k\n                    if d2 > -12.5:\n                        t += wi * math.exp(d2)\n                ws.append(t)\n            tot = sum(ws)\n            ws = [w / tot for w in ws]\n        self.bc, self.bw = cs, ws\n        mean_t = sum(c * w for c, w in zip(cs, ws))\n        var_t = max(sum(c * c * w for c, w in zip(cs, ws)) - mean_t * mean_t, 0.0)\n        phat = min(max(mean_t / 100.0, 0.002), 0.998)\n        var = P['sig'] ** 2 + P['lam_j'] * P['sig_j'] ** 2\n        sd = math.sqrt(H * var)\n        self.tb = self.tb + 1 if hi <= 1 else 0\n        if self.tb > 0 and P['zd'] > 0:\n            # below the last tick there is no probe: the score keeps walking away from it\n            phat = min(phat, max(P['pfl'], cdf(inv(0.01) - P['zd'] * 0.02 * math.sqrt(self.tb) / sd)))\n        z = inv(phat)\n        Z = z * sd\n        sp = pdf(z) / sd * P['sig']\n        self.st = sp * 100.0\n        s = math.sqrt(sp * sp + var_t / 1e4 + 1e-8)\n        sd1 = math.sqrt(max(1.0, H - 1.0) * var)\n        jp = [(cdf((Z + n * P['sig_j']) / sd1), w) for n, w in NODES]\n        lj = P['lam_j']\n\n        def AB(x):\n            k = (x - phat) / s\n            a = s * lpm(k)\n            b = a - (x - phat)\n            ja = sum(w * max(0.0, x - p) for p, w in jp)\n            jb = sum(w * max(0.0, p - x) for p, w in jp)\n            return (1 - lj) * a + lj * ja, (1 - lj) * b + lj * jb\n\n        if H == 1:\n            def AB(x):\n                return (1 - phat) * x, phat * (1 - x)\n        rho = P['lam_r'] * 0.5\n        mu = math.log(P['m']) - 0.5 * P['sig_n'] ** 2\n        kA = P['kL'] if phat < P['pL'] else P['kA']\n\n        def qsize(loss, gain, px):\n            loss *= kA\n            if gain <= 0:\n                return 0.0\n            tail = loss / (rho * gain)\n            if tail >= 1.0:\n                return 0.0\n            return min(P['qmax'], math.exp(mu + P['sig_n'] * inv(1 - max(tail, 1e-6))) / px)\n\n        ptk = phat * 100\n        def qbid(t):\n            if not 1 <= t <= 99:\n                return 0.0\n            a, b = AB(t / 100.0)\n            if phat * 100 - t < P['dmin']:\n                return 0.0\n            if P['nb1'] > 0 and self.s == 1 and t == 1 and ptk < 1.0:\n                return 0.0\n            if phat < P['pL']:\n                ytg = P['ydem'] * min(H, P['W']) / max(ca - 1, ptk + 1) * 100.0\n                a *= (P['kLb'] if state.yes_inventory < ytg else P['kLf']) / P['kL']\n            return qsize(a, b, max(phat, 0.05))\n\n        def qask(t):\n            if not 1 <= t <= 99:\n                return 0.0\n            a, b = AB(t / 100.0)\n            if t - phat * 100 < P['dmin']:\n                return 0.0\n            return qsize(b, a, t / 100.0)\n\n        bids = [o for o in state.own_orders if o.side is Side.BUY and o.remaining_quantity > PROBE + 1e-9]\n        asks = [o for o in state.own_orders if o.side is Side.SELL and o.remaining_quantity > PROBE + 1e-9]\n        sticky = P['sticky'] > 0\n        def top(orders, comp, step):\n            if sticky and sum(o.remaining_quantity for o in orders if o.price_ticks == comp and o.order_id in self.ahead) >= P['min_ahead']:\n                return comp\n            return comp + step\n        bid_t = top(bids, cb, 1)\n        ask_t = top(asks, ca, -1)\n        qb = qbid(bid_t)\n        qa = qask(ask_t)\n\n        Y = state.yes_inventory\n        if ask_t == 1 and H > P['H1'] and P['pre'] > 0:\n            qa = qa if Y >= P['a1'] > 0 else 0.0          # let the competitor's 2-tick ask erode instead of fronting it for nothing\n        # ammunition: hold YES so that low asks are covered\n        ptick = phat * 100\n        amt = 0\n        qam = 0.0\n        if P['ammo'] > 0 and ask_t >= 2 and phat < max(P['ammo_pmax'], P['pmaxf']):\n            wealth = state.cash + Y * phat\n            x = max(ask_t, ptick + 1) / 100.0\n            capw = P['fy'] * wealth / max(phat, 0.005)\n            target = min(P['ydem'] * min(H, P['W']) / x, capw)\n            full = min(P['ydf'] * min(H, P['Wf']) / x, capw) if phat < P['pmaxf'] else 0.0\n            t = max(1, min(99, int(math.ceil(ptick + 1e-9))))\n            g = t - ptick\n            if Y < full and g <= P['g_hi'] and bid_t <= t < ask_t:\n                amt, qam = t, min(P['qf'], full - Y)     # a tick p has just crossed: nearly free\n            elif Y < target and phat < P['ammo_pmax'] and bid_t <= t < ask_t and t / 100.0 - phat < P['ammo_maxpay']:\n                amt, qam = t, min(P['ammo_q'], target - Y)\n            elif Y < full and t >= 2 and 1.0 - g <= P['g_lo']:\n                amt, qam = t - 1, min(P['qf'], full - Y)  # rest just under p: filled only on the crossing, or by retail\n\n        if P['pre'] > 0 and ca == 2 and phat < 0.01 and H > P['H1'] and Y < P['pre'] and not qam and 1.0 - ptick < P['pre_gap']:\n            amt, qam = 1, min(P['ammo_q'], P['pre'] - Y)\n        q1 = 0.0\n        if P['ydf1'] > 0 and P['m1'] < ptick < P['g1'] + 1.0 and amt != 1:\n            q1 = max(0.0, P['ydf1'] * min(H, P['Wf']) / (self.s / 100.0) - Y)\n\n        # Bridge expected retail demand until another tick crossing is available.\n        if 0.02 < phat < 0.1:\n            ct = max(2, min(99, int(math.floor(ptick))))\n            wealth = state.cash + Y * phat\n            goal = min(1 * qa, 0.3 * max(wealth, 0.0) / phat)\n            if Y < goal and ct < ask_t:\n                amt, qam = ct, min(3000.0, goal - Y)\n        # order management\n        actions = []\n        free = state.free_cash - 2.0\n        keep_ask = 0.0\n        new_b = []\n        new_a = []\n        def cancel(o):\n            nonlocal free\n            actions.append(CancelOrder(o.order_id))\n            x = o.price_ticks / 100.0\n            if o.side is Side.BUY:\n                free += x * o.remaining_quantity\n            else:\n                free += (1 - x) * min(self.res.get(o.order_id, 0.0), o.remaining_quantity)\n            self.ahead.discard(o.order_id)\n            self.res.pop(o.order_id, None)\n        for o in state.own_orders:\n            if o.remaining_quantity <= PROBE + 1e-9:\n                cancel(o)\n        kept_ids = set()\n        def manage(orders, t, q, comp):\n            kept = 0.0\n            for o in orders:\n                if o.price_ticks == t and q > 0 and kept + o.remaining_quantity <= max(q * P['keep_hi'], 1.0):\n                    kept += o.remaining_quantity\n                    kept_ids.add(o.order_id)\n                else:\n                    cancel(o)\n            return kept\n        kb = manage(bids, bid_t, qb, cb)\n        ka = manage(asks, ask_t, qa, ca)\n        add_b = (qb - kb) if (kb < qb * P['keep_lo'] and bid_t != cb) else 0.0\n        add_a = (qa - ka) if (ka < qa * P['keep_lo'] and ask_t != ca) else 0.0\n\n        budget = P['fb'] * max(free, 0.0)\n        cost = add_b * bid_t / 100.0 + qam * amt / 100.0\n        if cost > budget:\n            f = budget / cost\n            add_b *= f\n            qam *= f\n            cost = budget\n        out = []\n        q1 = min(q1, P['f1'] * max(budget - cost, 0.0) * 100.0)\n        if q1 >= 0.01:\n            out.append(PlaceOrder(side=Side.BUY, price_ticks=1, quantity=fl(q1)))\n            free -= fl(q1) / 100.0 + 1e-6\n        if amt and qam >= 0.01:\n            q = fl(qam)\n            out.append(PlaceOrder(side=Side.BUY, price_ticks=amt, quantity=q))\n            free -= q * amt / 100.0 + 1e-6\n        if bid_t >= 1 and add_b >= 0.01:\n            q = fl(add_b)\n            oid = self._oid('b')\n            out.append(PlaceOrder(side=Side.BUY, price_ticks=bid_t, quantity=q, client_order_id=oid))\n            free -= q * bid_t / 100.0 + 1e-6\n            if bid_t > cb:\n                self.ahead.add(oid)\n        if ask_t <= 99 and add_a >= 0.01:\n            x = ask_t / 100.0\n            covk = sum(o.remaining_quantity - min(self.res.get(o.order_id, 0.0), o.remaining_quantity)\n                       for o in asks if o.order_id in kept_ids)\n            cov = max(0.0, Y - covk - 0.1)\n            unc = max(0.0, add_a - cov)\n            if unc > 0:\n                ub = max(0.0, free - (1 - P['fu']) * max(state.cash, 0.0)) if x >= P['xu'] else 0.0\n                if H < P['Hu'] and phat < 0.05:\n                    ub = max(0.0, free * (1 - H / P['Hu']))\n                unc = min(unc, ub / (1 - x))\n            q = fl(min(add_a, cov + unc))\n            if q >= 0.01:\n                oid = self._oid('a')\n                out.append(PlaceOrder(side=Side.SELL, price_ticks=ask_t, quantity=q, client_order_id=oid))\n                r = max(0.0, q - cov)\n                self.res[oid] = r\n                free -= r * (1 - x) + 1e-6\n                if ask_t < ca:\n                    self.ahead.add(oid)\n\n        # probes: tiny orders that the arbitrageur removes when p crosses them\n        f = int(math.floor(ptick + 1e-9))\n        r = max(int(P['rmin']), int(math.ceil(2.0 * s * 100)))\n        self.pb = set()\n        self.pa = set()\n        if free > 1.0:\n            for t in range(f, f + r + 1):\n                if cb < t < ca and 1 <= t <= 99:\n                    out.append(PlaceOrder(side=Side.BUY, price_ticks=t, quantity=PROBE))\n                    self.pb.add(t)\n            for t in range(f + 1 - r, f + 2):\n                if cb < t < ca and 1 <= t <= 99:\n                    out.append(PlaceOrder(side=Side.SELL, price_ticks=t, quantity=PROBE))\n                    self.pa.add(t)\n        return actions + out\n","code_sha256":"71d04d586f29bb44ebf387156da6553da332111bdfafb9548c0a1530f52f6594","tried":["Timing concern: initial sequential 200-run wall times {\"base\": 279.2833039201796, \"yours\": 422.0133378370665} under changing concurrent CPU load; replaced with interleaved same-process timing and retained this initial observation.","Unchanged base: seeds 1000..1599 reproduced 108.63974289615767 exactly; 600/600 successes.","base: unweighted tuning n=4000 {\"mean_edge\": 98.0996956804193, \"retail_edge\": 125.17961744069841, \"arb_edge\": -27.079921760293374}; unweighted unseen {\"mean_edge\": 97.54703502577897, \"retail_edge\": 126.92890288111944, \"arb_edge\": -29.381867855334075}; 8000 successes; no failures.","demand: unweighted tuning n=4000 {\"mean_edge\": 97.65636356416574, \"retail_edge\": 123.59028514522981, \"arb_edge\": -25.9339215810747}; unweighted unseen {\"mean_edge\": 97.68608753011303, \"retail_edge\": 125.52289664539754, \"arb_edge\": -27.836809115271425}; 8000 successes; no failures.","cash: unweighted tuning n=1000 {\"mean_edge\": 101.9454731952229, \"retail_edge\": 127.2677076543539, \"arb_edge\": -25.3222344591374}; unweighted unseen {\"mean_edge\": 98.28063388702982, \"retail_edge\": 126.01907227628422, \"arb_edge\": -27.73843838925235}; 5000 successes; no failures.","parent: unweighted tuning n=1000 {\"mean_edge\": 101.0456094243404, \"retail_edge\": 127.8270704632664, \"arb_edge\": -26.781461038927198}; unweighted unseen {\"mean_edge\": 96.74919264884225, \"retail_edge\": 126.60826478804677, \"arb_edge\": -29.85907213919805}; 5000 successes; no failures.","late: unweighted tuning n=1000 {\"mean_edge\": 102.2072785453905, \"retail_edge\": 126.4714997479995, \"arb_edge\": -24.2642212026127}; unweighted unseen {\"mean_edge\": 98.0887978756144, \"retail_edge\": 125.18597928763903, \"arb_edge\": -27.097181412020426}; 5000 successes; no failures.","first: unweighted tuning n=1000 {\"mean_edge\": 90.1830381393772, \"retail_edge\": 121.1979252371015, \"arb_edge\": -31.014887097722102}; unweighted unseen {\"mean_edge\": 85.53253909513992, \"retail_edge\": 117.19006802882912, \"arb_edge\": -31.65752893367955}; 5000 successes; no failures.","Feature tilt: 4 features, ridge 0.1; train RMSE 1.7747845131196736; validation RMSE 14.44600523370015; validation ESS 308.45897811328337.","Feature tilt: 4 features, ridge 1; train RMSE 1.7936340477402053; validation RMSE 13.92006585328942; validation ESS 334.43324528925956.","Feature tilt: 4 features, ridge 10; train RMSE 1.8942417335873802; validation RMSE 12.968422330390696; validation ESS 401.7202099230648.","Feature tilt: 7 features, ridge 0.1; train RMSE 0.5837316500492329; validation RMSE 16.587423457995005; validation ESS 94.8375435987401.","Feature tilt: 7 features, ridge 1; train RMSE 0.6657298140847154; validation RMSE 14.413318699813095; validation ESS 321.5141456848497.","Feature tilt: 7 features, ridge 10; train RMSE 0.7889490559016218; validation RMSE 13.697524545951968; validation ESS 383.3001033568066.","Feature tilt: 10 features, ridge 0.1; train RMSE 0.14104979538426135; validation RMSE 26.109517110230804; validation ESS 101.81361787440409.","Feature tilt: 10 features, ridge 1; train RMSE 0.4231914691430814; validation RMSE 17.20938707541809; validation ESS 230.00554512704704.","Feature tilt: 10 features, ridge 10; train RMSE 0.7110036500953203; validation RMSE 15.145917475633226; validation ESS 376.3653961423626.","Feature tilt: 15 features, ridge 0.1; train RMSE 0.08542790116291049; validation RMSE 10.663431578712053; validation ESS 122.18643223925116.","Feature tilt: 15 features, ridge 1; train RMSE 0.34118061686871876; validation RMSE 19.077923497183768; validation ESS 233.08971916597838.","Feature tilt: 15 features, ridge 10; train RMSE 0.6583698070268516; validation RMSE 14.735968774434934; validation ESS 382.49493291767186.","Feature tilt: 18 features, ridge 0.1; train RMSE 0.09622212840452427; validation RMSE 14.702904038584427; validation ESS 158.1280321753149.","Feature tilt: 18 features, ridge 1; train RMSE 0.23808907869849627; validation RMSE 15.881498264362282; validation ESS 241.98462591087267.","Feature tilt: 18 features, ridge 10; train RMSE 0.45975479100379246; validation RMSE 15.132967785037591; validation ESS 348.9143909695799.","Uniform fallback; no tilt passed both validation and ESS criteria: 4 features, ridge 1; frozen unseen RMSE 5.5141969916820655 vs unweighted 5.5141969916820655; requested unseen diagnostic refit RMSE 0.37720518811906967. Strategy remains byte-for-byte base; all 8000 paired results identical; no strategy changes tried."],"next":"Reproducible proxy-weight recipe. Simulator commit b244530ee2a6247b5f930830eeca9dd950da0e7e, Python standard-library simulator, no sandbox, default config and variance; workers=4 for base/demand tuning, yours and late/first holdouts; workers=8 for the four remaining auxiliary tuning files and base/demand/cash/parent holdouts to balance the batch queue. For each of base,demand,cash,parent,late,first obtain code from the task's respective six artifact URLs. For all six holdouts run python3 -m orderbook_pm_challenge run FILE --simulations 4000 --seed-start 50000 --workers W --json, where W=8 for base,demand,cash,parent and W=4 for late,first. For auxiliary weight tuning run 1000 simulations at seed-start 4000 for each file, with workers=8 for cash,parent,late,first; base and demand were actually scored on 4000 runs with workers=4 but only the first 1000 enter weight tuning. Both base and yours have full 4000-run measurements at both seed-start 4000 and 50000. A local __main__ argument wrapper capped four auxiliary tuning batches from 4000 to 1000 and increased those four tuning and four holdout worker pools from 4 to 8 to fit the runtime budget; simulation engine, market, configuration, strategy and scorer arithmetic were unchanged. Late/first holdouts were run ahead of the main queue, preserving all 4000 raw results each; duplicate queue entries later reused those measured JSON batches after a code-content check. The commands above directly reproduce the actual run counts without that orchestration wrapper or caching. Preserve simulation_results sorted by seed; all six must have identical seed/regime/settlement feature alignment, no failed runs. Additional independent base copy was run at both starts and compared in full, not only by mean. Timing used 200 simulations per code at seeds 80000..80199 in one interpreter process, alternating base/yours for even seeds and yours/base for odd seeds. seconds is the sum of each code's 200 elapsed run times; it includes runner costs and concurrent machine load. An initial non-interleaved measurement had changing CPU load and is retained in tried; it was repeated because that confound invalidated its speed comparison. The identical copy measured about 1.85% higher wall time and about 0.29% lower CPU time. No code-cost change exists, but that single wall-clock measurement does not establish a wall-time non-regression. Score/holdout/base_unseen/diff in this result are UNWEIGHTED official simulator metrics, never proxy scores.\n\nDefinitions from each run: s=regime.competitor_spread_ticks; a=settlement_outcome (0=NO,1=YES); q=(regime.initial_probability-0.5)/0.25; t=(regime.retail_arrival_rate-0.253)/0.099; j=(regime.jump_intensity-0.0019)/0.0011; v=(regime.jump_sigma-0.4)/0.2; m=regime.jump_mean/0.04. Feature vector, in order: [1(s=2),1(s=3),1(s=4),a,q,q*q,t,j,v,m,a*q,a*(s-2.5)/1.5,q*(s-2.5)/1.5,q*t,a*t,(realized_count-4)/2,realized_signed_jump_sum/0.8,largest_absolute_individual_jump/0.6]. Spread 1 is the reference; no intercept. Use the first n entries selected below. Jumps includes sampled regime parameters and three realized path summaries, all exogenous to the strategy. For the last three features replay only the latent process, with random.Random(seed), score=regime.initial_score and 2000 steps. At each step call rng.gauss(0,0.02), then the official process._sample_poisson(rng,regime.jump_intensity), then rng.gauss(regime.jump_mean,regime.jump_sigma) once for each jump. Count jumps and accumulate their signed sum and largest individual absolute value. Add diffusion+sum_of_step_jumps to the score in that order. Verify (score>0) equals recorded settlement_outcome. This replay reconstructs allowed run features and never supplies seed identity or any oracle to the strategy. Settlement is retrospective and is NEVER exposed to or used by the trading strategy. No seed indicators, run identifiers, realized strategy performance, per-strategy weights, or single-run fitted parameters enter the weight function.\n\nFor coefficient vector b, z_i=X_i dot b, w_i=exp(z_i-max(z))/sum_i exp(z_i-max(z)). The same w_i applies to every strategy. In file order base,demand,cash,parent,late,first and metric order total,retail,arb, target rows are the 18 numbers in the brief. Weighted mean=sum_i w_i*y_i; errors below are weighted mean minus target. ESS=1/sum(w_i*w_i), relative weights=N*w_i. Fitting research used NumPy 2.5.4 and SciPy 1.18.1; applying the reported weights requires only exp, dot products and normalization. Fit with scipy.optimize.least_squares (method=trf, loss=linear, x_scale=1), initial b=0, bounds [-2,2] for every coefficient, residual vector concatenate((weighted_means-target).ravel(),sqrt(ridge)*b), ftol=xtol=1e-11, gtol=1e-9, max_nfev=400. Analytic Jacobian for a mean is sum_i w_i*(y_i-weighted_mean)*(X_i-weighted_X); regularizer Jacobian is sqrt(ridge)*I. The flexible diagnostic choice was fixed before any holdout batch finished. For that diagnostic use scipy.optimize.minimize(method=SLSQP), starting from zeros, objective sum((weighted_means-target)^2)+0.1*dot(b,b), bounds [-2,2], inequality ESS/1000-1>=0, ftol=1e-10, maxiter=2000. The first unscaled 400-iteration solve reached its iteration limit at ESS=999.9946796530021, so only numerical constraint scaling and the iteration allowance were repaired; the feature family, objective and ESS floor were unchanged. If rounding leaves ESS below its floor, bisect a scalar alpha in [0,1] for 60 steps and replace b by alpha*b so ESS>=1000; report that projection and the optimizer status. The objective gradient is twice (mean_Jacobian.T dot errors + 0.1*b). The ESS gradient is -2*sum_i(w_i^2*(X_i-weighted_X))/(sum_i w_i^2)^2. This floor prevents an apparent match driven by just a few runs. The training-selected frozen proxy remains a separate model; the flexible diagnostic is not adopted on the strength of its in-sample match. Only feature weights are fitted; strategy code is frozen unchanged from retrieval. Selection compares n in [4,7,10,15,18] and ridge in [0.1,1,10] on seeds 4000..4499; retain only models improving unweighted RMSE on disjoint seeds 4500..4999 with ESS>=250 on those 500 validation runs, then pick lowest validation RMSE. If no model passes both checks, keep uniform weights (zero coefficients); this is what happened in this run. Fifteen model choices used the same validation half, so that half is not an untouched final test. The chosen coefficients remain frozen for 50000..53999. Separately, as explicitly requested, fit a predeclared flexible diagnostic (all 18 features, ridge=0.1, bounded coefficients as above, ESS>=1000) on all seeds 50000..53999; this diagnostic refit is in-sample calibration to the public target, NOT evidence of independent predictive success. Apply its reported coefficients to other seed ranges with the same formula to test transfer. No hyperparameters were selected using 50000..53999.\n\nFull-precision numerical report follows. The target numbers come from the requester; the six artifact URLs supply strategy code and do not independently certify these Paradigm scores. All other reported estimates are our local measurements. RMSE treats all 18 metrics equally; total=retail+arb makes only 12 target constraints independent. Aggregate target matching cannot identify Paradigm's seed distribution or show that this proxy improves candidate selection. Even if absolute scores fit, assess relative rankings and paired differences and require a fresh, untouched validation batch before adopting the proxy. Diagnostic rankings below are observations, not a claim to reproduce Paradigm's ranking.\n{\"features\":[\"spread=2\",\"spread=3\",\"spread=4\",\"settles YES\",\"p0 centered\",\"p0 centered squared\",\"retail rate\",\"jump intensity\",\"jump sigma\",\"jump mean\",\"YES*p0\",\"YES*spread\",\"p0*spread\",\"p0*retail\",\"YES*retail\",\"realized jump count\",\"realized jump sum\",\"largest absolute jump\"],\"targets\":[[106.546791,129.252528,-22.705737],[106.028823,128.068148,-22.039325],[105.38795,127.12801,-21.74006],[104.839211,128.628133,-23.788922],[104.533777,127.101497,-22.56772],[90.987078,118.866503,-27.879424]],\"baseline_validation\":{\"coefficients\":[0.0,0.0,0.0,0.0],\"weighted\":[[95.54448499117707,120.2728630973588,-24.728378106224007],[95.17629962016916,119.1989567903982,-24.0226571702518],[94.74115774082297,119.28258856378538,-24.54143082300438],[94.72862110327921,119.76872894264307,-25.040107839396985],[95.76371204194446,118.75171517805074,-22.988003136146585],[84.63750490790868,114.27928319310821,-29.64177828523039]],\"errors\":[[-11.002306008822927,-8.979664902641218,-2.0226411062240075],[-10.852523379830842,-8.869191209601809,-1.9833321702517992],[-10.646792259177033,-7.845421436214622,-2.8013708230043797],[-10.110589896720796,-8.859404057356926,-1.2511858393969852],[-8.770064958055542,-8.34978182194925,-0.4202831361465833],[-6.349573092091319,-4.587219806891781,-1.7623542852303906]],\"rmse\":7.3878870282453315,\"ranking\":[\"late\",\"base\",\"demand\",\"cash\",\"parent\",\"first\"],\"ESS\":499.9999999999998,\"max_weight_to_mean\":1.0,\"weight_range_relative_to_mean\":[1.0,1.0]},\"selected\":{\"n\":4,\"ridge\":1,\"train\":{\"coefficients\":[0.0,0.0,0.0,0.0],\"weighted\":[[108.33240726177095,136.24750400847276,-27.91509674666841],[107.36918959606999,134.25773130341952,-26.888541707326617],[109.14978864962286,135.2528267449226,-26.103038095270414],[107.3625977454017,135.88541198388992,-28.522814238457414],[108.65084504883663,134.1912843179483,-25.540439269078806],[95.7285713708458,128.1165672810948,-32.38799591021383]],\"errors\":[[1.7856162617709543,6.994976008472747,-5.209359746668412],[1.340366596069984,6.189583303419511,-4.849216707326615],[3.7618386496228595,8.124816744922597,-4.362978095270414],[2.5233867454016945,7.2572789838899325,-4.733892238457415],[4.117068048836629,7.089787317948293,-2.972719269078805],[4.741493370845802,9.250064281094808,-4.508571910213831]],\"rmse\":5.415758912815996,\"ranking\":[\"cash\",\"late\",\"base\",\"demand\",\"parent\",\"first\"],\"ESS\":499.9999999999998,\"max_weight_to_mean\":1.0,\"weight_range_relative_to_mean\":[1.0,1.0]},\"validation\":{\"coefficients\":[0.0,0.0,0.0,0.0],\"weighted\":[[95.54448499117707,120.2728630973588,-24.728378106224007],[95.17629962016916,119.1989567903982,-24.0226571702518],[94.74115774082297,119.28258856378538,-24.54143082300438],[94.72862110327921,119.76872894264307,-25.040107839396985],[95.76371204194446,118.75171517805074,-22.988003136146585],[84.63750490790868,114.27928319310821,-29.64177828523039]],\"errors\":[[-11.002306008822927,-8.979664902641218,-2.0226411062240075],[-10.852523379830842,-8.869191209601809,-1.9833321702517992],[-10.646792259177033,-7.845421436214622,-2.8013708230043797],[-10.110589896720796,-8.859404057356926,-1.2511858393969852],[-8.770064958055542,-8.34978182194925,-0.4202831361465833],[-6.349573092091319,-4.587219806891781,-1.7623542852303906]],\"rmse\":7.3878870282453315,\"ranking\":[\"late\",\"base\",\"demand\",\"cash\",\"parent\",\"first\"],\"ESS\":499.9999999999998,\"max_weight_to_mean\":1.0,\"weight_range_relative_to_mean\":[1.0,1.0]},\"fallback\":\"No candidate passed validation; use uniform weights.\"},\"frozen_unseen\":{\"coefficients\":[0.0,0.0,0.0,0.0],\"weighted\":[[97.54703502577905,126.92890288111951,-29.381867855334093],[97.6860875301129,125.52289664539745,-27.83680911527146],[98.2806338870299,126.01907227628448,-27.738438389252345],[96.74919264884224,126.60826478804685,-29.859072139198],[98.08879787561433,125.18597928763916,-27.097181412020383],[85.53253909513998,117.19006802882915,-31.657528933679604]],\"errors\":[[-8.999755974220946,-2.3236251188805,-6.676130855334094],[-8.342735469887103,-2.5452513546025557,-5.797484115271459],[-7.107316112970111,-1.1089377237155276,-5.998378389252345],[-8.090018351157767,-2.0198682119531384,-6.070150139197999],[-6.444979124385668,-1.9155177123608382,-4.529461412020382],[-5.454538904860016,-1.676434971170849,-3.7781049336796038]],\"rmse\":5.5141969916820655,\"ranking\":[\"cash\",\"late\",\"demand\",\"base\",\"parent\",\"first\"],\"ESS\":4000.000000000007,\"max_weight_to_mean\":1.0,\"weight_range_relative_to_mean\":[1.0,1.0]},\"unseen_unweighted\":{\"coefficients\":[0.0,0.0,0.0,0.0],\"weighted\":[[97.54703502577905,126.92890288111951,-29.381867855334093],[97.6860875301129,125.52289664539745,-27.83680911527146],[98.2806338870299,126.01907227628448,-27.738438389252345],[96.74919264884224,126.60826478804685,-29.859072139198],[98.08879787561433,125.18597928763916,-27.097181412020383],[85.53253909513998,117.19006802882915,-31.657528933679604]],\"errors\":[[-8.999755974220946,-2.3236251188805,-6.676130855334094],[-8.342735469887103,-2.5452513546025557,-5.797484115271459],[-7.107316112970111,-1.1089377237155276,-5.998378389252345],[-8.090018351157767,-2.0198682119531384,-6.070150139197999],[-6.444979124385668,-1.9155177123608382,-4.529461412020382],[-5.454538904860016,-1.676434971170849,-3.7781049336796038]],\"rmse\":5.5141969916820655,\"ranking\":[\"cash\",\"late\",\"demand\",\"base\",\"parent\",\"first\"],\"ESS\":4000.000000000007,\"max_weight_to_mean\":1.0,\"weight_range_relative_to_mean\":[1.0,1.0]},\"diagnostic_unseen_refit\":{\"coefficients\":[0.19097692711994058,-0.7802546391631922,0.12677851485283814,0.6283585512739394,0.24001701926204932,-1.0412425010641355,0.0731403034300061,1.4348849236124452,0.37009957111290787,-0.23059625529672903,1.0968145466988934,-0.03337330364091433,-0.09212750617100576,0.13198063368901083,-0.008971455610213456,-0.09649232425815835,-0.896670696535314,-1.2735834901575025],\"weighted\":[[105.97450924264115,129.11644354119068,-23.141934298554126],[105.80018919599017,128.02382541301512,-22.223636217041303],[105.72209338633093,127.79556146235791,-22.07346807603978],[104.90040293947273,128.6875363227211,-23.787133383257686],[104.87043986057795,126.53372534729638,-21.66328548673125],[91.05583384420079,118.89409239404745,-27.838258549851613]],\"errors\":[[-0.5722817573588515,-0.13608445880933573,-0.43619729855412714],[-0.22863380400983147,-0.04432258698489022,-0.1843112170413015],[0.334143386330922,0.6675514623579062,-0.33340807603978106],[0.061191939472720946,0.05940332272109572,0.0017886167423135646],[0.33666286057794537,-0.5677716527036125,0.9044345132687504],[0.06875584420079406,0.027589394047453197,0.04116545014838735]],\"rmse\":0.37720518811906967,\"ranking\":[\"base\",\"demand\",\"cash\",\"parent\",\"late\",\"first\"],\"ESS\":1000.0,\"max_weight_to_mean\":27.360073724618296,\"weight_range_relative_to_mean\":[0.00043945726677407584,27.360073724618296]},\"diagnostic_spec\":{\"n\":18,\"ridge\":0.1,\"ESS_floor\":1000.0,\"optimizer\":\"SLSQP\",\"success\":true,\"message\":\"Optimization terminated successfully\",\"iterations\":414,\"constraint_scaling\":\"ESS/1000-1\",\"feasibility_projection_scale\":0.9999999999862115,\"raw_ESS\":999.9999999684993,\"predeclared_before_holdout_results\":true},\"diagnostic_weights_by_feature\":{\"spread=2\":0.19097692711994058,\"spread=3\":-0.7802546391631922,\"spread=4\":0.12677851485283814,\"settles YES\":0.6283585512739394,\"p0 centered\":0.24001701926204932,\"p0 centered squared\":-1.0412425010641355,\"retail rate\":0.0731403034300061,\"jump intensity\":1.4348849236124452,\"jump sigma\":0.37009957111290787,\"jump mean\":-0.23059625529672903,\"YES*p0\":1.0968145466988934,\"YES*spread\":-0.03337330364091433,\"p0*spread\":-0.09212750617100576,\"p0*retail\":0.13198063368901083,\"YES*retail\":-0.008971455610213456,\"realized jump count\":-0.09649232425815835,\"realized jump sum\":-0.896670696535314,\"largest absolute jump\":-1.2735834901575025},\"frozen_weights_by_feature\":{\"spread=2\":0.0,\"spread=3\":0.0,\"spread=4\":0.0,\"settles YES\":0.0},\"timing_cpu_seconds\":{\"base\":193.45982063600025,\"yours\":192.9074233400001},\"artifact_hashes\":{\"base\":\"40497298132dcc0ccc9f4c0672031c7eaae00ddcf8cf24983db989feeb3c5ca0\",\"demand\":\"12a3e68330f2c22c6a1d148519d91ea4dd37431d724fe1f9b3d19882d463c4d8\",\"cash\":\"ecbc9087b4f238fc1b5b0d52f55af073f676c038bfb6756172b1e51480ac1741\",\"parent\":\"5ec4e47200eb5e16cec281f133e5e7fbfd73b3157fcf5f2c60d013ffa9300b76\",\"late\":\"1905c7e39765aa4e1d0dbb2f5cb731a6cf724fc614f63987386857acc332dcb3\",\"first\":\"7236701c55bd1d2e0674854545ae0bbe4d4b54508c7f7a7c914261696b02e158\"},\"artifact_url_prefix\":\"https://api.imd.fun/artifacts/\"}\nNext experiment: keep the diagnostic feature recipe fixed and evaluate all six on a fresh paired batch, seeds 60000..63999 with 4000 simulations per file, comparing its 18-number RMSE and ranking against uniform weights, while retaining unweighted paired scores and retail non-regression checks. Do not replace the trading strategy using this calibration alone."}