**commit** e580f9d125e93911d67fe76e80798ffe3faa428b
**parent** 9f13bc4362f0763aec047ec7a0608453503b022d
**Author:** Sebastiano Tronto <sebastiano.tronto@gmail.com>
**Date:** Tue, 28 Dec 2021 16:05:43 +0100
fixed typo
**Diffstat:**

1 file changed, 1 insertion(+), 1 deletion(-)

**diff --git a/README.md b/README.md**
@@ -135,7 +135,7 @@ pruning table to solving EO with HTM moveset and this coordinate would have valu
The values for most pruning tables are memorized modulo 16, so they only occupy
4 bits per entry, and values larger than 15 are saved as 15. This is good enough
for most applications.
-Some large tables are memorized in compact form using only 4 bits, similarly
+Some large tables are memorized in compact form using only 2 bits, similarly
to what [nxopt](https://github.com/rokicki/cube20src/blob/master/nxopt.md) does:
a base value `b` is picked and a value of `n` is saved as `MIN(3,MAX(0,n-b))`.
When a value of `v=1,2,3` is read it is simply returned as `v+b`, while if