5m + 15m, Kalshi + Polymarket US.
Every call is hash-chained before the contract settles. Misses stay in the ledger. No proven edge yet, and we say so.
Recompute the chain yourself:
https://t.co/rZCc53rH0Y
Not financial advice.
@0xDataWolf polymarket's own post carves 5m/15m btc out of the 0.01 taker cut on https://t.co/IhnMZz59nW, so the short btc windows miss the big discount. a 5m backtest that skipped fees was already wrong. it stays wrong after friday.
Not financial advice.
@earias41 "the first number you see is usually wrong" is the whole job. same with resolve rates: a side settling yes more often only matters against the price paid at entry, fees included.
we log our kalshi 15m btc calls before the close, misses included.
Not financial advice.
kalshi 15m btc, nov 17 utc, 2 min out, n=79 of 96:
both sided (n=74): us 70 right, 4 wrong. market 70 right, 4 wrong.
brier, all 79: us 0.065, market 0.062. market beat us.
worst miss: row 66830. we said 96.9% up. it settled down.
https://t.co/rZCc53rH0Y
Not financial advice.
@KlayBrowning589 a 15m bitcoin window can look like a print until you keep the misses in the same ledger. the close doesn't care how sneaky the path was.
Not financial advice.
@Millon_Sancet a new wrapper on 5-minute btc up/down doesn't change the part that usually gets skipped: the touch size at send time, and writing the miss when the close disagrees.
Not financial advice.
@JeremyKissel@Kalshi 0dte still has hours. a 15m btc up/down window can reprice inside the time it takes to confirm the fill. the useful record is every window, misses included, written before the close.
kalshi 15m btc, nov 16 utc, 2 min out, n=88 of 96:
both sided (n=84): us 76 right, 8 wrong. market 76 right, 8 wrong.
brier, all 88: us 0.066, market 0.063. market beat us.
worst miss: row 65969. we said 85.0% down. it settled up.
https://t.co/rZCc53rH0Y
Not financial advice.
kalshi 15m btc, nov 15 utc, 2 min out, n=93 of 96, hash-chained:
both sided (n=86): us 78 right, 8 wrong. market 78 right, 8 wrong.
brier, all 93: us 0.082, market 0.084.
worst miss: row 65726. we said 98.9% up. it settled down.
https://t.co/rZCc53rH0Y
Not financial advice.
new: a polymarket us 5m btc desk.
our P(up) next to the book's, four models side by side, top of book and tape. it's in the paid tier.
free: every kalshi 15m call, hash-chained before settlement.
https://t.co/fFsWXeaxv9
Not financial advice.
@whizzkee stripping unrealistic fills is the right first cut. the test that matters is grading all 417 at the real ask plus fees, misses included, so the $888 isn't mid-price paper. still a long way from a proven edge, as you said. NFA
kalshi 15m btc, nov 11 utc, 2 min out:
us 71 right, 6 wrong. market price 70 right, 8 wrong.
brier: us 0.091, market 0.088. market still better.
worst miss: row 63042. we said 6.9% up. it settled up.
hash-chained pre-settlement: https://t.co/rZCc53r9bq
Not financial advice.
5m callers, from their receipts: Nov 7–11 UTC (partial Nov 11).
@TerryCallsBTC posted receipts for 68 numbered Polymarket US 5m calls: 58 won, 10 lost, by his own grading. We tallied what he posted; we did not re-check those wins and losses against CF Benchmarks settlement.
Average entry 84.8¢; equal shares would need an 84.8% win rate to break even before fees. His sample hit 85.3%.
Not financial advice.
@JohannMarkov Agree on the feedback loop. It only teaches you something if the misses can't be quietly dropped. That's why we log every published Kalshi 15m BTC forecast to a public hash chain and score it against the market's own quote, misses included.
A miss from this morning's Kalshi 15m BTC ledger (KXBTC15M-26NOV101000-00):
11 min before close we had UP at 9.9%. The market had 10.5%. It settled UP. Brier 0.811, ledger row 62453.
That row stays. It was timestamped before the block settled, and each settled row is sealed into a sha256 hash chain, so a quiet edit would break every hash after it. Recompute the chain in your browser: https://t.co/4LM2wI0I2J
Experimental, no proven edge. Not financial advice.
Win rate is a bad scoreboard for up/down contracts. A share bought at 75¢ has to win 3 times in 4 just to break even (before fees), so a high hit rate can still lose money.
The fairer test is calibration: score every call against the price at the time. We do that with Brier on our Kalshi 15m ledger. Current result: the market's live quote out-Briers every model we've registered, at nearly every offset. That miss is on the page: https://t.co/kDBgJclMUf
Experimental, no proven edge. Not financial advice.
@w3motion good teardown. the other half of the audit is the wallets nobody posts. a public record only means something if the misses stay in it. (that's how we run our Kalshi 15m ledger.) Not financial advice.
Polymarket US 5m BTC vs Kalshi 15m BTC: same question (up or down?), different clocks. One opens every 5 minutes, the other every 15.
Our public ledger covers Kalshi KXBTC15M only: each call is hash-chained before the block closes, and misses stay in. Check it yourself: https://t.co/4LM2wI0adb
Experimental, no proven edge. Not financial advice.
@kaeose For the BTC 5m accounts now losing, I'd separate entry prices, fees and position size from win rate. A run of winning windows can still lose money; a wallet's full trade history matters more than its best streak. Not financial advice.
Polymarket US Up/Down doesn't settle on the last tick. Open and close are each the average of the 60 CF Benchmarks BRTI prints in the minute before (per Polymarket US docs). A wick in the final seconds moves that average less than the chart suggests.
Our public ledger is Kalshi 15m: every call hash-chained before settlement, misses included. Study floor: 20 UTC days; until it clears, it reads "not yet edge."
Free ledger + verify: https://t.co/cFhDibPKbB
Not financial advice.