What's a centipawn?

After using computer analyse there is something that says average centipawn loss. Can anyone explain what that a centipawn is?

is it 1/100 of a pawn?

1/10th

Mistaken, it is 1/100. Engine way of analysis, eg 0.79 evaluation of a position is +0.79 centipawn advantage for white

Its essentially a measure of how many centipawns you effectively failed to win by making less than the best moves in the position. The lower the centipawns the better your play.

Thanks everyone

More specifically, if we assume that computers play optimally (or do so in 99% of positions), then any move that differs from the computer’s choice is suboptimal.  If you are White and the best move should give an evaluation of 0.7 and your move is evaluated at 0.4, then you have lost 0.3 centipawns with that move.  Or, in other words, your move was a third of a pawn worse than best play.

If you take every such move and average them out, you get your average centipawn loss.  In theory, if you play perfectly, your every move matches the computer choice and your centipawn loss is 0.  The more mistakes, the higher the number.  You can use your average centipawn loss as another guide to chess improvement.  That is, as you get better, this number should get lower, similar to golf.

The number means very little in just one game, but taken as an average over many games, it can be useful.  If you know your normal centipawn loss is about 60 for instance, and you run a string of games under 50, you know you are playing better, even if you draw or lose some of those games.