Every week Polymarket runs price ladders on the biggest stocks in the market. Will NVIDIA touch $212 this week? Will Tesla drop to $405? Each level is its own contract. If the stock trades through the price during the week it pays out, and if it doesn't, it goes to zero.
That makes these markets easy to grade. So we graded all of them. We pulled every weekly ladder Polymarket has run on AAPL, AMZN, GOOGL, META, MSFT, NVDA and TSLA, rebuilt the trading history of all 1,764 contracts from 38,581 individual trades, and checked every confident quote against 15-minute candles of the actual stock.
Overall, when the market put more than 58% odds on a price target, the stock went on to hit it 72.6% of the time. That average gets a lot more useful once you split it by how confident the market actually was.
The market is honest above 70%
Here's every call sorted by what the market was charging when the signal fired, next to how often the stock actually got there:
| Market was quoting | Calls | Stock got there | Move needed |
|---|---|---|---|
| 58% to 65% | 57 | 56.1% | 2.51% |
| 65% to 70% | 28 | 67.9% | 1.88% |
| 70% to 80% | 45 | 80.0% | 1.40% |
| 80% to 90% | 19 | 84.2% | 1.51% |
| 90% and above | 26 | 92.3% | 1.52% |
A 65 to 70% quote hit 67.9% of the time. An 80 to 90% quote hit 84.2%. Above 90%, 92.3%. Top to bottom, the price tracks reality. The traders in these markets aren't bluffing.
The weak spot is the cheap end. Calls priced between 58 and 65% only hit 56.1% of the time. You're paying 60 cents for something that's barely better than a coin flip. The information starts around 70%. Below 65%, you're mostly buying noise.
When it hits, it runs
A hit rate doesn't tell you what the move was worth, so split it the other way: how far away was the target when the call fired, and how far did the stock actually go?
| Outcome | Calls | Move needed | Move realized |
|---|---|---|---|
| Stock reached the target | 127 | 1.33% | 4.10% |
| Stock fell short | 48 | 3.29% | 1.19% |
When these calls hit, the stock didn't touch the target and turn around. It needed 1.33% on average and ran 4.10%, about three times the distance. The misses went nowhere: 1.19% delivered on a 3.29% ask.
What's doing the predicting there is distance, not the odds. Targets inside about 1.3% got hit, and targets past 3% mostly didn't, whatever the market was charging for them.
How we measured it
For each contract we walked the trade history in time order and took the first trade after Wednesday that crossed 58%. We pinned that moment to the open of the 15-minute candle covering it, then scanned every candle through Friday's close. If any candle's high reached the level (or its low, for downside targets), that's a hit. Polymarket's own settlement never enters into it, only the candles do.
We also stripped out everything that would pad the results. We grade from the first crossing instead of the peak, because these markets settle the moment the stock touches the level, so every winner ends at 100% and the peak just tells you who won. If the stock had already gapped past the target before our entry candle, that call was never a prediction, so 55 of those got cut. And there's a class of five and six unit trades priced at 99 cents that shows up at identical timestamps across unrelated tickers. Those aren't real quotes, so we set a 25-unit minimum trade size and cleared them out.
The full dataset
We're publishing the entire dataset with this article, one row per call, so you can rebuild every number yourself:
- FINAL_58_signals.csv All 376 signals: the quote and size of the trade that fired it, the crossing time, the entry candle and its open, the target, the move needed, whether the stock got there, and how far it ran.
- VERIFY_signals.csv 315 signals lined up against the live Polymarket page, one URL and settled outcome per row, for spot checking.
- weekly_rungs_scored.csv All 4,140 weekly contracts with their resolution and the week's actual high, low and close.
- hit_trades.json The raw 38,581 Polymarket trades behind everything, timestamped.
To land on the same 175 rows as the tables above: drop every row where pct_needed is zero or negative, then every row where signal_size is under 25. The cascade is 376 raw, 321 after removing gap-throughs, 175 after removing dust trades. Hit rates: 73.1%, 68.5%, 72.6%.
A few real rows so you know what you're looking at:
| Ticker | Week of | Target | Quote | Needed | Hit? | Stock ran |
|---|---|---|---|---|---|---|
| AMZN | Jul 27 | ↑ $240 | 59% | 2.85% | Yes | 17.09% |
| META | Jul 27 | ↓ $580 | 76% | 2.23% | Yes | 11.59% |
| TSLA | Apr 13 | ↑ $375 | 97% | 1.21% | Yes | 10.46% |
| TSLA | Apr 20 | ↓ $367.50 | 63% | 2.07% | No | 1.84% |
| AAPL | Apr 27 | ↑ $292 | 63% | 9.14% | No | 7.35% |
| MSFT | Apr 27 | ↑ $450 | 58% | 9.54% | No | 1.53% |
Look at the AAPL row. The market wanted a 9.14% move in two and a half days and charged 63 cents for it. The stock ran 7.35%, a monster week, and still fell short.
Every market since launch
These ladders launched on March 27, 2026, and we tested every single one that has settled since: 18 weeks, 126 events, 1,764 contracts, and every confident call inside them. That's 175 qualifying signals across AAPL, AMZN, GOOGL, META, MSFT, NVDA and TSLA, 98 upside and 77 downside. Nothing cherry-picked, no survivorship, the whole record.
So if you actually trade these markets, here's how we'd play it. Trust the ladder above 70%. Skip anything under 65%. And check the distance before you check the odds, because a target inside 1.3% is probably getting hit and a target past 3% probably isn't, whatever the market says.