diff --git a/doc/zstd_compression_format.md b/doc/zstd_compression_format.md
index cd7308de196..0532a846f45 100644
--- a/doc/zstd_compression_format.md
+++ b/doc/zstd_compression_format.md
@@ -1083,7 +1083,8 @@ It depends on :
   Presuming an `Accuracy_Log` of 8,
   and presuming 100 probabilities points have already been distributed,
   the decoder may read any value from `0` to `256 - 100 + 1 == 157` (inclusive).
-  Therefore, it must read `log2sup(157) == 8` bits.
+  Therefore, it may read up to `log2sup(157) == 8` bits, where `log2sup(N)`
+  is the smallest integer `T` that satisfies `(1 << T) > N`.
 
 - Value decoded : small values use 1 less bit :
   __example__ :