Why L5, L10 and L20 disagree — and which one to trust

Two people watch the same player, look at the same statistic, and come away with different numbers. One says he is hitting the line about sixty per cent of the time. The other says forty. Both are reading real counts, and neither is lying. They are using different windows.

This is the most common disagreement in esports statistics, and it is not a disagreement about the player. It is a disagreement about which maps count as evidence.

One line, counted three ways

Here is a hypothetical line — one player, one threshold, twenty maps played. The counts do not change. Only the window does.

All three rows are arithmetically true. The 5-map window is not more accurate than the 20-map window, and the 20-map window is not more current than the 5. They answer different questions, and the number in the final column only means something once you know which question was asked.

Why a longer window is calmer

Sample size controls how much a single map can move the answer. Take the same player and change the result of one map in each window:

One close map in a five-map window turns 40% into 60%. That is not a form surge; it is one map. This is the whole reason a small sample is untrustworthy — not that it is wrong, but that it is unstable enough that the next map is likely to move it a lot.

Sample size is also not the only thing that can shrink a window. A rate is only as current as the maps the database actually holds, which is why ClutchIQ prints recent series coverage next to a form record: seven stored matches out of the last twenty is a narrower slice of the calendar than the same seven wins suggests.

Why a shorter window is not automatically 'current'

“Recent form” is a phrase worth distrusting. A short window is only more current if something actually changed — a roster move, a role change, a map pool shift, a patch. If nothing changed, a short window is just a noisier read of the same underlying rate.

• Something changed on the roster or in the role? The short window is carrying real information, and the long window is averaging across a break.
• Nothing changed, and the short window disagrees? Most of that gap is noise. Widen the window before drawing a conclusion.
• You cannot tell which it is? Say so. That is the honest answer, and it is more useful than a confident one.

Match the window to the match

The window should match the conditions of the match you are looking at, not some general preference for recent or long-run numbers.

• Same lineup, same patch, same map pool as the maps you are counting? The window is comparable and its length is a trade between calm and currency.
• A stand-in is playing, or the AWPer changed? The older maps describe a team that no longer exists. Narrow until the roster is constant, and accept the smaller sample.
• A map rework landed mid-window? Maps before it are a different map. This is not a technicality — map win rates shift after a patch, and averaging across the change hides the shift you are trying to see.

The same reasoning applies to LAN versus online, which is a window filter of a different kind. Splitting a sample by venue is the same decision as splitting it by date: both trade size for comparability.

The threshold is the other window

Window length is the visible half of the problem. The threshold moves the number just as much. A 20+ kills line and a 22.5+ kills line are different questions about the same player in the same maps, and one can be a comfortable hit rate while the other is a coin flip.

So a fully stated rate needs three things attached: the statistic, the threshold, and the window. “He hits about sixty per cent” is missing two of them.

What to do with a number you disagree with

When someone quotes a rate you do not recognise, the productive move is not to argue about the rate. Ask which maps it counted, over which threshold. Most disagreements of this kind resolve as soon as the window is named, because both numbers were true.

ClutchIQ shows the window and the sample size beside every rate for exactly this reason. If you want to see what a price demands before you go looking at form, the break-even calculator does that arithmetic on any price in one step.