Research
Essential question: does how much a bucket's price has risen during the day correlate with how likely it is to actually win - independent of what time that rise happened? Every strategy on this site so far enters on a price level (is it priced 50-70% right now?) or a fixed clock time (7am, 2pm, and so on). Delta asks a different question: does the size of the move itself - a bucket that jumped 15 points versus one that crept up 2 - carry information about the eventual outcome, regardless of when in the day that move happened? If it does, that's an entry signal that reacts to the market itself instead of the clock.
For each city-day: find the unique top-priced bucket at the 6am-local snapshot (that city's own real local time, no starting price floor or ceiling) - call this the baseline. Track that same bucket's price forward to the snapshot closest to 6am + the duration selected below, and compute the change in points ("degree delta", not a relative % change). Bin by delta, and compare the actual win rate of that bucket in each band. Built from every Polymarket snapshot captured since 2026-05-22, through 2026-11-06.
At the default duration, all cities: 955 qualifying city-days spread across 16 delta bands, averaging about 59.7 per band - 10 of 16 bands are still below the 30-sample minimum used elsewhere on this site. Read the shape as suggestive, not confirmed.
At the current accumulation rate (~5.68 qualifying city-days/day across all cities), this sample would roughly double by 2027-04-23, averaging ~119.4/band - worth a second look then to see whether this pattern is holding up or was partly noise.
Does the result change depending on entry price? The baseline here has no price floor (see Method above) - a 6am leader at 25% is currently pooled the same as one at 70%. Worth splitting the sample by baseline price band to check whether the delta/PnL relationship found here holds, strengthens, or breaks down depending on how confident the market already was at 6am.
X-axis is the change in the bucket's own price, in points, from the 6am-local baseline to the selected duration later (band midpoint, 5-point bands - negative means the price fell). The dashed line is the overall win rate across every delta, positive or negative.
Research
Essential question: does the size of a bucket's price move, combined with how long since baseline, predict how profitable buying it would have been? Rows are duration since baseline (1-9 hours), columns are the price delta since baseline (positive moves only, 2-point bands to 40 points), and color is realized PnL per unit net of fees - one grid per baseline hour, 6am through 2pm local. Trust a smooth gradient that holds up across several neighboring cells and repeats across multiple baseline-hour grids, not an isolated bright cell surrounded by pale ones - that's usually 1-2 lucky or unlucky city-days, not signal. Individual cells need 30+ samples to count and most don't clear that yet, so read this by grouping neighboring cells - a whole delta column, a duration row, or a few adjacent bands - rather than trusting any single cell alone.
Every cell from all 9 baseline-hour heatmaps below, pooled by delta band alone (summed across baseline hour and duration) to get a real sample size per band - individual heatmap cells average only about 10 samples each, too thin to trust on their own. Win rate (left axis) should climb steadily as the move gets bigger; watch whether avg PnL per unit (right axis) climbs with it or peaks early and fades - that gap is the market re-pricing the bucket about as fast as its odds improve, which is where any real edge would have to live.
Every cell from all 9 heatmaps above that clears the 30-sample minimum and shows a positive avg PnL per unit. This is the reliable-and-winning subset only - a short list here isn't bad news, it just means most of the grid is still too thin to trust one way or the other (see the sample-size note above). Click a column header to sort by it.
| Baseline | Duration | Delta Band | Samples | Win Rate | Avg PnL/Unit |
|---|---|---|---|---|---|
| Loading... | |||||