Typing speed data from real duels

Almost every typing statistic online describes a solo test. This one describes races. The figures below come from completed TypeBag duels — two people typing the same passage at the same time, unable to restart, with a stake on the result. As far as we know nobody else publishes that, because almost nobody else runs it.

What this is not

It is not the average typing speed. It cannot be, and we would rather say so than let the numbers be quoted as though it were.

TypeBag's players chose to look for competitive typing and then chose to put money on it. They are not a cross-section of anyone: they are faster than the general population by selection, before a single key is pressed. This is exactly the problem we describe on the average typing speed page — a site reporting its own visitors' figures is describing people who came looking, not people in general — and publishing our numbers as a population average would be us doing the thing we criticised.

What these figures do describe, honestly, is what typing looks like under competitive conditions. That is a narrower claim and a rarer one.

Method

Every figure here is computed directly from completed duels, recounted daily. Words per minute is correct characters divided by five, divided by elapsed minutes, which is the same arithmetic the WPM calculator and the typing speed test use. Four categories of data are deliberately left out:

• Practice runs. The character counts and timings for a solo practice session are reported by the browser. The server does the arithmetic, but on numbers the client supplied, so they are not evidence of anything.
• Ghost duels. Racing a recorded run saves a row for the recording as well as the challenger, so a popular ghost would be counted many times over. Those rows are also written with a perfect accuracy value rather than a measured one, which would quietly pull the accuracy figures upward.
• Free-tier bots. The synthetic opponents that fill an empty free queue have a speed chosen for them by a governor. It is a setting, not a measurement, and no bot ever produces a typing record at all.
• Unfinished races. A disconnect or an abandoned duel leaves a partial run whose speed describes a race that stopped, not a pace anyone typed at.

What remains is matched and private duels between real accounts, where every keystroke was validated by the game server as it arrived. Those are the only numbers on this site that a player could not simply have told us, and they are the only ones we are willing to publish.

Limitations worth stating

• The passages are continuous English prose of roughly 30 to 60 seconds, with normal capitals and punctuation. A test made of common lowercase words would produce higher numbers from the same hands.
• Mistakes must be corrected before a race can finish, so these speeds already include the cost of fixing errors. Tests that let you type past a mistake report faster times.
• Paid duels are matched by rating, so most people race opponents near their own speed. That affects who plays whom; it does not affect how fast any individual typed.
• These are aggregates over all players. Nothing here identifies a person, and no individual figures are published.

Using these figures

You are welcome to cite or reproduce them with attribution to TypeBag and a link to this page. If you do, please carry the qualification with the number: these are competitive duel figures, not population averages. A statistic quoted without its method is how the unsourced numbers we complain about got into circulation in the first place.

The most direct way to find out where you sit in this distribution is to enter it. The free tier needs no account and no deposit — race someone, or read how a duel works first.