Edge decay — why your sweet spot stops working
Every edge you find has a half-life. The angle that produced two profitable seasons will stop producing, and the uncomfortable part is that it usually stops quietly — the results just get worse, slowly, while everything about your process feels identical. This lesson is about telling the difference between an edge that died and an edge that is having a bad month, and what to do in each case.
Why edges die
Four mechanisms, and they look nearly identical from the outside.
The book fixed its model. Pricing models get rebuilt. A book that was systematically slow on backup-player props hires someone, buys a feed, or ships a new version, and a gap that existed for eighteen months closes in a week. This is the cleanest death: it happens at one book, at a specific moment, and it is often visible in the data as a step change rather than a drift.
Other sharps found the same gap. The more common case. An angle that is discoverable by you is discoverable by others, and when enough money arrives on the same side, the price moves before you get there. Your edge does not vanish — it gets absorbed into the opening line. This is the mechanism behind almost every angle that "used to work."
Your account got limited. This one is invisible in your win rate and brutal in your returns. You still find the edge, you just cannot get size on it. A $50 maximum on a market where you used to get $2,000 is a dead edge in every way that matters, and it does not show up anywhere in a spreadsheet of results.
The market matured. Structural, not adversarial. A market with thin liquidity and slow pricing attracts volume, volume attracts attention, and five years later it prices like a major market. Props broadly have gone through this. So has most of the low-hanging arbitrage that existed in the first years of US legalization.
The diagnostic: CLV, not results
Here is the problem with using results to detect decay. Results are almost useless over the samples you actually have.
Take a bettor whose true edge produces a +2% ROI at -110 pricing — a genuinely good, sustainable number. Over 200 bets, the standard error on observed ROI is about 6.7 percentage points. The 95% range of what that bettor should expect to observe runs from about −11.2% to +15.2%.
Read that again. A real, durable +2% edge routinely shows up as a double-digit loss over 200 bets. If you are diagnosing decay from a results curve, you will fire every good angle you have and keep several bad ones.
Closing line value settles it, and it settles it fast. CLV compares the price you took against the market's final price. It is a measure of whether you were ahead of the market at the moment you acted, and it converges in dozens of bets rather than thousands, because it does not have to wait for outcomes.
So the diagnosis is a two-by-two:
CLV holding, results lagging. This is variance. Your prices are still ahead of the close, which means the market keeps agreeing with you after the fact. Nothing is wrong. Do nothing except keep sizing correctly and keep logging.
CLV degrading, results anything. This is decay. You are no longer getting the price before the market does. It does not matter whether this month's results were good — a winning month with negative CLV is a warning, not a reward.
CLV holding, results excellent. Working as intended, and still not a reason to size up beyond your rules.
CLV degrading, results excellent. The most dangerous quadrant in betting, because it feels like the best one. You are getting paid for being lucky in a market that has already moved past you.
Track CLV as a rolling average by angle and by market, not as one blended number for your whole account. A blended CLV hides exactly the thing you are looking for: one specific angle going flat while everything else holds it up.
Reading the shape of the decay
Once CLV says decay, the shape tells you the cause.
A step change — CLV fine through a specific week, then flat afterward — points at a model rebuild or a limit. Check your maximum stake at that book first, then check whether the change is at one book or all of them. One book means they fixed something. All books means the market moved.
A slow drift over months points at competition or maturation. Your average CLV on the angle goes from +2.2 points to +1.4 to +0.6. Nobody did anything to you; there is just more money on your side of the trade than there used to be.
Fine on small bets, gone on large ones is a limit problem wearing a decay costume. Your edge is intact and your access is not.
The response
Three moves, in increasing order of effort.
Rotate markets. The fastest response and often enough. The skill that found the edge is transferable; the specific market is not. A bettor who was good at NBA backup-player props because they were fast on news is good at NHL goalie-confirmation windows for the same reason. Rotation is also the single best defense against account limits, because concentrated action in one market is what gets flagged.
Go deeper into less-covered surfaces. Books allocate pricing attention to handle. That allocation is uneven and always will be, which means there is always a frontier. Smaller conferences, second-tier props, first-half and quarter markets, and any league where the slate is too large to price carefully — college basketball in January is the canonical example — will price worse than the marquee board for structural reasons that will not be fixed.
Rebuild the angle. The slowest and most durable. If a signal stopped working, something about the underlying mechanism changed. Backtest it again against past seasons, split the sample by season, and find where it broke. Sometimes the angle is still real but needs a tighter filter — it worked on road teams and never really worked at home, and the aggregate was carrying the losing half for years. That is not decay, that is an angle you never fully understood, and it is recoverable.
While you rebuild, size honestly. An angle you are unsure about gets quarter-Kelly, not half. The 5% bankroll cap does not move for anything, including confidence.
The part nobody wants to write down
No edge is permanent. Not one. The books are staffed, funded, and motivated, and the market gets more efficient every year. Anyone telling you they have found something durable is describing something they have not measured over a long enough window.
What is durable is the process: shopping every number, measuring CLV on every bet, sizing under a cap, and rotating before you get limited rather than after. That process does not decay. It is what lets you find the next edge before the current one finishes dying, which is the actual job.
The bettors who last are not the ones who found a great angle. They are the ones who noticed, early and without flinching, that it had stopped being one.