162 wallets tracked. 31 of them are right 85%+ of the time. Here’s how we find them — and what they’re doing right now.
Prediction markets aggregate thousands of bets into a single number — the odds. But that number mixes everything together: a tourist dropping $50, a bot arbitraging pennies, and a wallet that has been right on 94% of 200+ markets.
The odds you see on Polymarket or Kalshi are an average. An average of noise and signal, mixed together. You can’t tell which is which. That’s the problem we solve.
We started simple: follow wallets across hundreds of resolved markets — politics, crypto, geopolitics, tariffs. Record every position, every entry price, every outcome.
Out of thousands, a small group keeps winning. Not by luck — across hundreds of markets, over months. We call them alpha wallets.
When an alpha wallet takes a position, we see it. We know their accuracy, track record, and which side they’re on.
Right now, 14 alpha wallets are active on the Iran regime market. Two are Tier-1. All 14 are on NO. Their combined position: $582K.
This isn’t a guess. This is data.
View wallet leaderboard →Polymarket says 72%. Kalshi says 58%. Same event, different prices. One of them is wrong.
We compare prices across 7,900+ markets. When we find a gap, we check: are alpha wallets on one side? How many? What’s their accuracy on similar markets?
See current arbitrage gaps →One accurate wallet on a position — interesting. Five on the same side — that’s a signal.
We measure conviction: how concentrated are the best wallets on one side? When multiple high-accuracy wallets independently choose the same direction, the probability of them all being wrong drops fast.
We publish our signals in advance. When markets resolve, we show the receipts.
We publish a weekly Accuracy Scorecard. Every signal, every outcome, including the misses. See full scorecard →
Alpha wallets are the foundation. But a wallet moving doesn’t tell you the full story. Is there news driving the trade? Do the numbers actually support it? What if five different AI models looked at the same market — would they agree?
That’s what our signal engine does. Three layers, working together.
We don’t trust a single model. Every market runs through five independent AI analysts — each with a different approach:
A fundamentals analyst that updates probabilities as new evidence arrives. A quant that looks at volume patterns and historical base rates. A contrarian that asks what the market is NOT pricing in. A cross-validator that weighs all evidence equally. And a macro analyst that tracks global context.
When four out of five agree — that’s conviction. When they split — the market is genuinely uncertain, and we say so.
We call the result an AI Verdict — a single consensus score that tells you what five independent models think this market is actually worth.
See AI verdicts on live markets →| Model | Says | Confidence |
|---|---|---|
| Claude Analysis | NO | 92% |
| DeepSeek Quant | NO | 88% |
| Grok Contrarian | NO | 85% |
| Gemini Validation | NO | 90% |
| Kimi Macro | NO | 87% |
| CONSENSUS: 5/5 NO — Strong conviction · Combined fair price: 8¢ YES (market: 9.5¢) | ||
Markets don’t move in a vacuum. A sanctions announcement, a ceasefire rumor, an election result — these change probabilities in hours.
We pull live news for every active market, every six hours. When Iran’s foreign minister makes a statement at 2 AM, our next pipeline run connects it to every Iran-related market and recalculates.
The AI models don’t just see wallet data. They see context. That’s the difference between “price dropped 5 cents” and “price dropped because the State Department confirmed new sanctions.”
View latest market intelligence →We don’t say “this looks good.” We calculate it.
Every signal has a statistical backbone. A Bayesian model starts with the market price and updates it based on how accurate the wallets have been historically. A second model measures how much of the trading volume comes from informed versus retail money. A third combines wallet consensus, position sizes, and historical accuracy into a single compound score.
The result: a fair price estimate. When the market says 62 cents and our models say 85 cents — that’s a measurable edge, not an opinion.
162 wallets monitored. Tier-1 consensus calculated. Five AI models run. Live news pulled. Three math models computed. Cross-platform gaps checked.
All of that collapses into a single card: the Verdict.
It tells you the direction — BUY YES or BUY NO. It shows you where to enter, where the targets are, and where the math says to stop. One view. No digging through dashboards. No assembling signals yourself.
This is what you actually see when you open any market on OddsShift:
5 tier-1 wallets unanimous YES. Math models confirm: fair value 84c. News: White House confirmed tariff takes effect as scheduled.
Under the hood: 162 wallets, 5 AI models, live news from Reuters/Bloomberg/WSJ, 3 math models (Bayesian, PIN, Compound Signal), and cross-platform gap analysis. All running every 6 hours. All compressed into this one card.
Every number here is traceable. The wallet count — on-chain data. The AI models — five independent runs. The targets — statistical calculations with confidence scores. Nothing is a black box.
Most tools show what Polymarket already shows for free — in a different layout. Some track whale trades. Some show charts. Some send alerts.
None answer the only question that matters: why are the odds moving?
“The price moved from 72¢ to 58¢”
“The price is at 72c. Our Verdict: BUY YES. 5 AI models agree. 13 wallets with 87% accuracy confirm. Math says fair value is 84c. News supports the thesis. Target: +36% yield.”
| Feature | Other tools | OddsShift |
|---|---|---|
| Raw odds | ✓ | ✓ |
| Whale alerts | some | ✓ |
| Cross-platform gaps | rare | ✓ |
| Wallet accuracy tiers | ✗ | ✓ (162) |
| AI Models (5 independent) | ✗ | ✓ |
| Live news intelligence | ✗ | ✓ |
| Math signal layer | ✗ | ✓ |
| AI Verdict (one view) | ✗ | ✓ |
| Public accuracy data | ✗ | ✓ |
| Honesty about limits | ✗ | ✓ |
We could show only the wins. Every other tool does that. Here’s the truth:
Alpha wallets aren’t always right. 87% on tariffs is exceptional. On Federal Reserve markets, accuracy drops to 27%. Smart money isn’t smart about everything.
AI models disagree sometimes. When our five models split 3-2, that’s not a strong signal — it’s genuine uncertainty. We label it as such. No false confidence.
We don’t know the future. Signals are based on what historically accurate wallets are doing now, validated by AI and math. They’ve been right far more often than not. But markets are uncertain.
We’re transparent about bugs and limits. We publish our accuracy weekly — including the misses. If we find a data issue, we disclose it.
Why say this? Because trust built on honesty lasts. If you want guaranteed profits, we’re not it. If you want to see what the smartest money in prediction markets is actually doing — with AI analysis, live news context, and statistical proof — that’s us.
| Category | Signals | Correct | Rate | |
|---|---|---|---|---|
| Tariffs | 12 | 10 | 83% | ✓ |
| Geopolitics | 8 | 6 | 75% | ✓ |
| Crypto | 5 | 3 | 60% | ~ |
| Fed / Economics | 4 | 1 | 25% | ✗ |
We also built xTweet Market — real-time tweet predictions that outperformed Polymarket crowds with 3% median error. It’s one of many tools powered by our signal engine.
Plans from $29/mo. Full access to every signal, every market, every AI Verdict. Decide for yourself if it’s worth knowing.