Convexly

Skill or luck? We check whether a Polymarket wallet's edge is real, with a confidence interval and the method public. Independent, no venue money.

[Convexly @convexlyapp]

https://t.co/CAEdurpBKu

[Convexly @convexlyapp]

Sample read so you can see the format, run fresh today. A wallet from our published top-50 board, "Zarvantis" 0x30c9...8770:

35 wins / 12 losses on 47 resolved positions. That is a 74% win rate. Realized edge: -2.5pp (95% CI [-7.9pp, +1.9pp], n=47 resolved positions). The interval includes zero. Verdict: "Not separable from luck."

A 74% win rate is not edge if the entry prices implied more than 74%. We read entry price vs resolution, not W/L.

Full record: https://t.co/NeQQap3VKl

Past resolved record. Not a forecast, not advice.

[Convexly @convexlyapp]

If you trade on @Polymarket, drop your wallet address below and you’ll get a free skill-vs-luck read of your resolved record: realized edge, 95% CI, plain verdict. Published methodology, same test for every wallet.

Fair warning: of 3,871 wallets with 30+ resolved positions, 178 cleared the luck filter. Most reads won’t come back “Skilled on the record.” That’s the point.

Diagnostics, not advice. Not a forecast. Subtract the luck.

[Convexly @convexlyapp]

Read the clause before calling a lobbyist. It bans extensions that facilitate real-money transactions on predictive outcomes. Same line regulators draw: execution is the regulated part. The policy page adds "or promote," so analytics extensions sit in a gray zone, but websites are untouched. Anything that lives as a buy button inside someone else's store just learned what platform risk means.

[Convexly @convexlyapp]

@DextersSolab The 95% win rate trick is just buying 95-cent favorites over and over. You're right 19 times, then the 20th blows up and eats all 19. Win rate stays gorgeous, the account bleeds. Most gameable number on the site.

[Convexly @convexlyapp]

@PredMarketWiki Thank you @PredMarketWiki team! Excited to join the community.

[PM.WIKI @PredMarketWiki]

🆕 New project added to the prediction markets wiki

Convexly

@convexlyapp helps traders evaluate whether a Polymarket wallet or market is worth trusting 📊🔍

It scores wallet skill and market quality, making it easier to identify reliable signals before you trade or cite them

[Convexly @convexlyapp]

New lane we just opened: we measure how concentrated the trading in a market is, then ask whether crowded-into markets resolve cleaner than diffuse ones.

We pre-registered the test before looking. 0/40 resolved in the concentrated stratum, 0/40 in the diffuse one. Both still accruing.

If you think concentration is a quality signal, you're guessing. We wrote down the bet first.

[Convexly @convexlyapp]

@PolyDekos Selling tails at 80c prints every week until it doesn't. One "lock" resolves no and you hand back two months in an afternoon. These threads only ever show the green weeks. Post the full curve.

[Convexly @convexlyapp]

Most edge claims never let you check them.

Ours checks itself. Our wallet-skill test pits 178 candidate wallets against 3,693 controls, frozen since May. It needs 40 qualifying wallets to even run. Today it sits at 38. Two away.

We can't touch the rule. It reads "not enough data yet" until the data clears the bar. The honest answer, on schedule.

[Convexly @convexlyapp]

Everyone quotes the odds a Polymarket market shows. Almost no one tracks whether the market resolved cleanly: voided, UMA-disputed, no terminal price.

We now score resolution quality per market and tied it to a pre-registered forward test. 4 of 30 markets resolved so far. The bar fills on outcomes, not opinion.

A clean number on a market that pays out ugly is not a clean number.

[Convexly @convexlyapp]

Our public Polymarket Top-50 has 12 wallets that are net-negative on realized money. We left them in on purpose.

A board you can't lose a spot on for losing isn't a board. It's a billboard.

If a leaderboard never shows you its losers, ask what else it's hiding.

[Convexly @convexlyapp]

We ranked our own top-50 Polymarket cohort by skill, not profit.

0 of the 35 testable wallets clear our bar. The other 15 of 50 were too thin to test.

"Not separable from chance" isn't "no skill." A PnL screenshot just can't tell them apart.

https://t.co/WcWgOtaMYM

[Convexly @convexlyapp]

We filed four strictly-prospective forward tests on our own work. Today all four return the same pre-committed answer: not enough data yet.

That is the honest read, written before we looked. No human can override it. They mature on time and outcomes alone.

Watch us pass or fail in public.

[Convexly @convexlyapp]

Everyone wants to copy the top Polymarket wallet.
Most of the time you'd be inheriting one lucky bet, not a skill.
So we re-scored the leaderboard by whether each wallet's record can be told apart from luck. Most can't.
Check any wallet before you copy it 👇
https://t.co/xnP25Bz8BW

[Convexly @convexlyapp]

@thenarrator Would add a fourth: resolution. Catalogue and incentives get traders in the door, but one badly settled market burns more trust than a thin book ever will, and it never shows up in a dashboard until it's terminal. Have you seen a venue actually die from settlement disputes yet?

[Convexly @convexlyapp]

@DataBasedBets @Polymarket The glitch isn't even the scary part, every venue ships bugs. The scary part is a settlement dispute coming down to screenshots versus the support desk, with no neutral record to check. Did the order history from the API match what you bought, or did that flip too?

[Convexly @convexlyapp]

@PolyDekos Buying the heavy favorite side thousands of times is the classic high win rate trade. The win count measures discipline, the tail grades the strategy: one freak weather week across a few of those cities and the ledger reads very differently. Has anyone pulled his worst drawdown?

[Convexly @convexlyapp]

@Angello3400PRMR @Polymarket Honest question: how many exact scores in a row before it stops being a hot morning and starts being skill? Two is where every legendary trader story begins, and most of them end the same way.

[Convexly @convexlyapp]

If you sell wallet signals, a dare: run your own wallet through https://t.co/zXVm4kZ6pY and post the result, whatever it says. We ran our skill read on our own published top-50 cohort. 35 of 50 had enough resolved positions to test. Zero cleared the corrected statistical bar.