The pivot that found our wedge: from a generic decision tool to a skill-vs-luck audit for Polymarket wallets
I spent six months building a generic decision-intelligence platform. Then I looked hard at how people actually used it and made a call: the interesting problem was somewhere else.
Prediction-market traders already carry the one thing most products lack, a resolved track record, and they have a real question nobody answers well: is my P&L skill, or is it luck? So I pivoted Convexly to read a Polymarket wallet and score exactly that.
What that turned into:
- A free wallet analyzer (convexly.app/tools/polymarket-wallet-analyzer). Paste an address, get a calibration read in about 30 seconds, no signup.
- A resolution pipeline that only scores settled positions, so the read reflects outcomes, not open hope.
- A published calibration audit of the top Polymarket profit wallets.
The part I am proudest of is not the launch, it is what happened after. We published that audit, then caught a methodology error in our own analysis: we had scored calibration per orderbook fill instead of per deduplicated position, which inflated the reported numbers. We corrected the post, regenerated the charts, and updated the public dataset. The directional finding held (worse calibration tracks with higher raw profit, Spearman r = +0.42), just smaller than first reported, and we said so.
That is the bar I want to hold Convexly to: frozen methods, public data, and negative results that stay up even when they are inconvenient.
If you build somewhere your own numbers can be wrong, I would genuinely like to compare notes on catching and correcting in public. And if you follow prediction markets, the analyzer is free to try.
Really solid breakdown. The part that stands out most is how quickly you moved from “interesting engagement” to a much clearer wedge. My guess is the biggest win here is letting users get a useful result before asking for signup, because that usually makes the upgrade path feel earned instead of forced.
I 100% agree. The feedback that was consitent from all the first users was that the gated information was a definite turnoff. Being able to give them value before even signing up is already bringing more views and users. So hopefully the pivot turns this trickle into a stream, so to speak.
“This is one of the cleanest pivot breakdowns I’ve seen — especially catching your own methodology error publicly. That builds real trust.
Highest-leverage change here feels like removing the signup wall and letting the value prove itself first.
You should test this in a live competition. $19 entry, winner gets a Tokyo trip (flights + hotel).
Round 01 just opened (100 cap) — best odds right now.”
That's an awesome idea! Thank you for sharing!
Thanks Convexly! Glad the pivot breakdown resonated.
Quick overview of Tokyo Lore: It’s a paid ideas competition where people submit Tokyo-connected business or creative ideas. For $19 you get a custom AI-generated artifact of your idea + a full SPEAR business analysis, plus entry into the round where the winner gets a real trip to Tokyo (flights + hotel booked by us).
Prize pool has started building — odds are excellent right now while it’s still very early.
Would you be interested in submitting an idea? Happy to send you the direct $19 link and walk you through the whole process (very quick).
What do you think?
That’s a really interesting result — especially with 1,500+ responses but no paid users.
Usually when this happens, it’s not the traffic, but how the value is presented after the first interaction.
Something in the flow might be creating friction or confusion before users reach the payment step.
Happy to share a quick audit if helpful.
Thank you for the feedback! I’ll take that into consideration