Convex lake

historical orderbooks, trades & research for polymarket, kalshi, limitless, + crypto options/perps. free for students. 10% off via referral ↓

[Convex lake @convexlake]

We rebuilt a real account's equity curve as if copytraded. Leader's real result: +$45,492 across 600 resolved markets. Same trades, with only the copytrader's fill replaced by a realistic delay: -$16,163 to -$29,146. Same edge, wiped out.

[Convex lake @convexlake]

@Masum4990 @sight_hood early exit is underrated. a recent paper argues clob market makers earn mostly by carrying mispriced positions to settlement, not from the spread. that's why amm and clob designs keep coexisting

[Convex lake @convexlake]

@armouredme a "breaking" book on a niche market is often just empty levels. on one kalshi btc event i found 91% of strikes parked at the minimum tick, defaults not quotes. check last trade time before reading the price as a view

https://t.co/NIvcG4V8vE

[Convex lake @convexlake]

@denitsa_tsekova @nmasc_ valuation aside, catalog depth is still lopsided. last count: polymarket 578 live crypto markets vs kalshi 43, weather 568 vs 3. the capital race is about closing that gap

https://t.co/nZcug24WzL

[Convex lake @convexlake]

read the settlement rule, not the headline. last shutdown kalshi settled off a morning snapshot of opm's site, polymarket needed an announcement by 11:59pm et. polymarket's nov 12 contract hit 97¢, then went to ~1¢ at midnight. same event, different question

https://t.co/3AyPT1hhCV

[Convex lake @convexlake]

@Brokx_is_back careful reading volume when an airdrop is on the table. one study found polymarket's naive election volume overstated real turnover by more than 2x. farmed volume inflates the leaderboard, not the depth

[Convex lake @convexlake]

@Langerius @Outcomexyz the interface was the bottleneck, agreed. nice side effect: hip-4 outcome books live in hyperliquid's core order book, so you can study them like perps. bank deposits should bring very different flow than crypto natives

[Convex lake @convexlake]

@Yasmina_sm @Outcomexyz real growth. one thing to separate: volume in a few markets vs spread across them. on kalshi only 12% of the btc strike ladder ever saw a trade. listings ≠ liquidity

https://t.co/6AkkGNOcz2

[Convex lake @convexlake]

@VikingoDigital_ volume ath is nice, but depth is the number i'd watch. 24h volume says nothing about how much you can fill at the touch. that's what tells you if the growth is real

[Convex lake @convexlake]

@carsonthedev the real difference is the gate: listing is permissionless, but behind a hype stake. polymarket and kalshi curate instead. for me the question is depth, not listings. i've been snapshotting hip-4 orderbooks since october and that's where it'll show

[Convex lake @convexlake]

@KalshiSports fun one: sports is the worst-calibrated category in my polymarket data, ~0.16–0.19 brier vs ~0.05 for tech. a 15% favorite this far out mostly means nobody knows yet

[Convex lake @convexlake]

@bckfv_eth sports is where this makes sense. i computed polymarket calibration by category: sports brier ~0.16–0.19 vs ~0.05 for tech/entertainment, and it barely improves toward resolution. more mispricing to work with, if you can size like that

https://t.co/qfZViYEdPy

[Convex lake @convexlake]

@SheikhSilicon @Polymarket buying cheap longshots in size is usually the losing side of the favorite-longshot bias. research on 300k+ kalshi contracts points to longshots being overpriced. unless it hedges something else, that's paying the crowd

[Convex lake @convexlake]

@NgocNguyen21127 B, not close. 7/7 happens by luck all the time. the big polymarket study (1.72m accounts) found only ~3% of traders show skill beyond chance. 200 trades at 70% is a track record, 7 is a streak

https://t.co/piGNCe41SD

[Convex lake @convexlake]

@fere_ai the scan is the easy part. the trap is resolution: two venues can list the "same" headline with different deadlines or sources. i'd want the agent reading the rule text, not the title

[Convex lake @convexlake]

right. official docs still say 250ms taker delay on fast crypto markets. there are reports it was cut to 50ms for 5m/15m, not confirmed in the docs yet. and nobody publishes real submit-to-fill latency, only ping to a cdn edge. that's the number that matters for stale quotes

https://t.co/cR5Qf0rvr8

[Convex lake @convexlake]

@Primee32 51% at half a cent per dollar is thinner than the taker fee at 50¢. so it's mostly maker fills or rebates doing the work. that's the interesting part of the run, more than the win rate

[Convex lake @convexlake]

@Dan1ro0 38 trades an hour on btc up/down means fees and fills decide everything, not the signal. at 50¢ a taker can pay up to ~3.5% on polymarket crypto markets. i'd want to know how much of that pnl is maker side before anyone copies it

[Convex lake @convexlake]

@coleodds @PolyScalping 85k predictions from a bot is exactly the account you can't copy. its edge lives in the fill. when i replayed a profitable arb account with copier latency, +$45k became -$16k. stats sites should show strategy type next to pnl

[Convex lake @convexlake]

@chefcryptoz @bagel_win one-tap copy is great ux, but the leader's fill isn't yours. on fast crypto markets the book can empty in seconds. i saw a $0.02 fill turn into $0.995 for a same-size copy 10s later. worth showing users their expected slippage next to the leader's pnl

[Convex lake @convexlake]

@TheeAlphaWolf1 @pots_market the part people skip: you copy the trade, not the fill. i replayed a real limitless arb account as a copier. leader +$45k, copier -$16k to -$29k, same trades. a directional account survived with a ~5% haircut. pick who you copy by strategy type, not pnl

https://t.co/tqUGpJlpCy