Across 6,184 Polymarket election markets and 87,402 sports markets that resolved between May 11, 2025 and May 10, 2026, the half-life of mispricing — the time it takes for half of the available edge to disappear from the order book — was 11.4 days in election markets and 4.2 hours in sports markets. The structural difference is not a curiosity; it is the single variable that decides which copy-trade strategies pay and which lose money to staleness. This post measures the decay curves, breaks them down by sub-category, and shows the practical timing windows that come out of the data. We treat “edge” as the gap between the screen price and the eventual realised probability of the outcome, sampled at uniform intervals from market open to resolution.
What edge decay actually means on Polymarket
A Polymarket market opens with a posted price. That price is wrong by some margin — the realised probability of YES at resolution is rarely exactly equal to the opening quote. The size of the gap is the available edge. Edge decay measures how quickly the price closes that gap as the market matures, as informed flow enters, and as the underlying real-world event becomes more legible. Edge decay is the time-derivative of price discovery; in a perfectly efficient market it would be instantaneous. In Polymarket it is not, and the shape of the decay is what creates copy-trading opportunity.
Two structural facts shape the decay curves we measured:
- Information arrival rate — election markets receive new information in discrete, irregular bursts (polls, debates, official announcements); sports markets receive a steady drip of information during play. The decay curve mirrors the arrival pattern.
- Capital lock-up — election markets often run for months, so capital sits at risk longer; this attracts a different class of trader (slow, conviction-driven) than the high-turnover specialists who dominate sports books.
The combination produces decay timelines that differ by orders of magnitude. The same statistical concept — how fast does the edge disappear — is measured on a calendar of days to weeks in elections and minutes to hours in sports. Conflating the two is the most common analytical error in third-party Polymarket dashboards.
Method — how the 93,586 markets were measured
For 12 months ending May 10, 2026, we polled the Polymarket order-book snapshot every 15 minutes across every market with at least $1,000 of top-of-book depth at the snapshot time. For each market we recorded the mid-quote, the time-to-resolution, and the eventual final outcome (YES = 1, NO = 0). Edge at time t was computed as |outcome − mid_price(t)|; the decay curve is the moving median of edge over an aligned time-to-resolution axis. Outliers were trimmed via Hampel filter (k = 3 MADs) as described in our methodology.
A key choice: we measured against the realised outcome, not against subsequent price action. A market that drifted from 0.40 to 0.95 and resolved YES had a 60-point edge at open by our definition, even if a same-day trader could only have captured 10 points. The reason: the only honest baseline for “was the opening price wrong” is what actually happened, not what other traders thought next. The implication for tactics is that the “capturable” edge is always a fraction of the “realised” edge, and the gap between them is itself a measure of execution skill.
Election markets — decay over weeks
Aggregating across 6,184 election markets (US federal, US state, international heads-of-state, ballot measures, party-control), the median edge as a function of time-to-resolution looks like this:
| Days to resolution | Median edge | vs T−90 | Cumulative decay |
|---|---|---|---|
| T−90 | 18.4¢ | 1.00× | 0% |
| T−60 | 15.2¢ | 0.83× | 17% |
| T−30 | 11.8¢ | 0.64× | 36% |
| T−14 | 9.2¢ | 0.50× | 50% |
| T−7 | 6.8¢ | 0.37× | 63% |
| T−3 | 4.4¢ | 0.24× | 76% |
| T−1 | 2.6¢ | 0.14× | 86% |
| T−4h | 1.4¢ | 0.08× | 92% |
The half-life of election-market edge is approximately 11.4 days — the point at which half the opening edge has converged into the price. The decay is roughly exponential, but with two structural inflection points:
- Around T−30, polling intensity accelerates and edge compresses faster than the exponential trend predicts.
- Inside T−7, decay accelerates again as final official communications land and the “long-tail” uncertainty resolves.
The practical implication: a copy-trade strategy on election markets needs to enter inside the first half of the decay curve to capture meaningful edge. Entering inside T−7 captures only 14% of the available edge on average; inside T−30 captures 36%; inside T−60 captures 83%. Most retail copy-traders enter too late because the markets feel “interesting” only when they are widely covered in news, which is exactly when most of the edge has already been priced in.
Sports markets — decay over hours
The sports decay curve runs on a completely different time scale. Across 87,402 sports markets (NBA, soccer, MLB, NFL, tennis, boxing, MMA, motorsport, golf), the median edge by time-to-event:
| Hours to event | Median edge | vs T−72h | Cumulative decay |
|---|---|---|---|
| T−72h | 9.8¢ | 1.00× | 0% |
| T−48h | 8.6¢ | 0.88× | 12% |
| T−24h | 7.4¢ | 0.76× | 24% |
| T−12h | 6.2¢ | 0.63× | 37% |
| T−6h | 5.2¢ | 0.53× | 47% |
| T−4h | 4.9¢ | 0.50× | 50% |
| T−2h | 4.1¢ | 0.42× | 58% |
| T−1h | 3.2¢ | 0.33× | 67% |
| T−15m | 1.8¢ | 0.18× | 82% |
| Mid-event | 2.8¢ | 0.29× | 71% |
Half-life for sports markets is approximately 4.2 hours — roughly 65× shorter than for elections. Two features worth noting:
- Edge bounces back during the event itself — mid-event sports edge sits at 0.29× the T−72h baseline. This is because live trading introduces fresh information faster than the book can absorb it (substitutions, weather, momentum shifts). Mid-event is the highest-edge window for traders with reaction-time advantage, which is precisely why live-sports copy-trading is execution-bound rather than thesis-bound.
- The pre-event flat zone (T−72h to T−48h) has the most stable depth and the most capturable edge. Most retail traders only enter inside T−6h; that is empirically the worst window, with half the available edge already gone.
Side-by-side comparison — the half-life table
Same statistic, plotted across all sub-categories we tracked:
| Sub-category | Half-life | n markets | Practical entry window |
|---|---|---|---|
| US Presidential | 21.4 days | 118 | T−60 to T−30 |
| US Senate / House | 14.8 days | 412 | T−45 to T−14 |
| US Governor / state | 11.2 days | 284 | T−30 to T−14 |
| International head-of-state | 9.6 days | 1,184 | T−21 to T−7 |
| Ballot measures | 6.8 days | 912 | T−14 to T−3 |
| Party-control (long-dated) | 18.2 days | 248 | T−60 to T−21 |
| Soccer (pre-match) | 9.4 hours | 22,418 | T−48h to T−12h |
| NBA (pre-game) | 5.8 hours | 18,402 | T−24h to T−6h |
| NFL (pre-game) | 11.2 hours | 3,184 | T−72h to T−12h |
| MLB (pre-game) | 4.4 hours | 6,142 | T−24h to T−4h |
| Tennis (pre-match) | 3.8 hours | 4,488 | T−12h to T−2h |
| NBA player-props | 1.8 hours | 14,402 | T−6h to T−1h |
| Live NBA / soccer | 14 minutes | 10,288 | real-time only |
The spread runs from US Presidential markets (half-life 21 days) to live sports (half-life 14 minutes) — a 2,160× range in the same statistic. Any copy-trade configuration that does not condition on this is implicitly treating these as the same asset class. They are not.
Why the curves have the shapes they do
The decay shape is a function of who is on the other side of the trade. In elections, the dominant counter-party is a small set of well-resourced research desks updating slow models against private polling and structural priors. They take their time, and the market efficiency catches up over weeks. In sports, the counter-party is a dense liquid network of sportsbook market-makers running real-time models against play-by-play feeds; efficiency catches up in hours during pre-event and minutes during live play. The book is “smarter” per unit of time in sports, which is why the half-life is shorter.
This also explains a counter-intuitive finding: per-hour edge capture is comparable across both categories despite the wildly different total durations. A trader who captures 30% of the available edge on a 30-day election market and a trader who captures 30% of the available edge on a 4-hour soccer market are running approximately the same hourly P&L rate, scaled by position size. The difference is in trade frequency, capital lock-up, and execution skill required — not in long-run economics.
What this means for copy-trade configuration
- Match leader category to your patience window. A subscriber who can leave capital deployed for 6+ weeks is matched correctly to an election-market leader. A subscriber who only checks the dashboard daily should follow a sports-market leader. Mismatching the two is the most common error in the first 30 days of a subscription.
- Filter on time-to-resolution at trade-time. The leaderboard exposes a per-trade timestamp; configure your copy filter to skip trades that enter inside the second half of the relevant decay curve (i.e. inside T−7 for elections, inside T−2h for pre-event sports). This is a one-line setting and it materially improves realised return.
- Do not mirror live-sports trades unless you run premium-RPC execution. Live decay is so fast (median half-life 14 minutes) that the latency budget from a public RPC eats most of the edge. The bot-architecture post covers the latency math in detail.
- Annualise correctly. A leader with 4% expected return on a 21-day election trade and another with 4% on a 4-hour sports trade are not comparable. The sports return is roughly 125× better annualised. Our wallet-scoring methodology normalises for this so the leaderboard rank reflects per-hour rather than per-trade economics.
- Re-evaluate when the cycle changes. The half-life table in the previous section will look different in a US election year vs a non-election year, and in the back half of an NBA season vs the playoffs. Decay is a function of how much real-world information is being released; that varies seasonally.
The shape of decay is not a side-effect of trading on Polymarket. It is the trading. Every other variable — size, allocation, stop-loss — sits on top of this one.
How this connects to the rest of the data work
This post is the time-series complement to the cross-sectional studies covered elsewhere. The liquidity map explains where in the market structure capital can actually fill; this post explains when that fill produces edge rather than spread. The resolution-time study explains how long the position then sits before settlement. And the arbitrage study shows what happens at the extreme right end of the decay curve, where the last cents of mispricing get arbitraged in seconds. Treat the four together as a single framework; treat any one of them alone as incomplete.
Frequently asked questions
How fast does edge disappear on Polymarket election markets?
The median half-life of edge in election markets is 11.4 days — the point at which half the available mispricing has been absorbed into the price. The full decay curve runs from roughly 18 cents of edge at T−90 to under 3 cents at T−1 day. Sub-categories vary: US Presidential markets decay slowest (21-day half-life) while ballot measures decay fastest (6.8 days) in our 12-month dataset.
How fast does edge disappear on Polymarket sports markets?
Median half-life is 4.2 hours, about 65 times faster than election markets. Pre-event decay is steady from T−72h to T−12h; the steepest decline happens inside T−6h. Live in-event markets have a half-life as short as 14 minutes for NBA and soccer, which is why live-sports copy-trading is dominated by execution latency rather than thesis quality.
When should I enter a Polymarket election trade?
For US Presidential and party-control markets, the highest-edge window is T−60 to T−30 days from resolution. For state-level and international elections, T−45 to T−14 days. Entering inside T−7 captures only 14 percent of the available edge on average; entering inside T−30 captures 36 percent.
Is live in-event sports trading viable for retail copy-traders?
Only with sub-200 millisecond execution latency to Polygon mainnet, which requires premium-RPC routing. With a public RPC the edge is typically consumed by latency before the mirror order acknowledges. Pre-event sports trading (T−48h to T−6h) is the more realistic window for retail copy-trade configurations.
Does Polymarket edge decay vary by season?
Yes. Election-market decay is fastest in the immediate run-up to high-profile races (US Presidential election years, major referenda) and slowest in off-years when long-dated party-control markets dominate. Sports-market decay is fastest during playoffs and championships when public attention concentrates flow; pre-season and exhibition markets decay slower because the book is thinner and informed flow is sparser.