2/2/13 UPDATE

I have finished collecting and analyzing the next test, where I managed to obtain ~3000 data points of the same condition: Sivir attacking tank Mundo, with a single zeal and no runes/masteries to influence the 10% crit chance.

TL;DR Version is below.

__Final Results For 10% Critical Chance__

With this test, I collected 3436 points whether Sivir's autoattack crit or not. For each normal attack, a value of 0 was stored in the Excel document. For each critical, a value of 1 was recorded. Each attack also had the time stamp included on the data.

After the test concluded, this test had a final chance of crit of 8.55%. Unlike the previous test, this one did have several critical attacks early, pushing the early percentage above 10%. However, it quickly dropped to around 8% and remained there, even dipping as low as 7%.

Attached are now comparisons of the first 600 attacks, as well as the 2nd set of data and a combined check as the final graph. I believe that the assumption I can attach the second set of data to the first to get a cumulative probability is valid since the conditions of the test were as identical as I could make, and the only difference should be the random function driving the critical attack chance.

__What does this mean?__

What this means is that any game lasting beyond 500 attacks seems to have an artificially low critical chance. Whether that is a function of buying new items or just total attacks is something I might look into.

My next test will be 4% crit, and then 50%. I might also come back to look at crit chance as a function of when it is bought to when it drops off, as buying new crit chance items might reset their function, changing it away from the artificial low.

TL;DR

3000 points later, 8.6% for 10% Zeal.

Attacking over 300 times a game makes you artificially low.

5000+ points of data combined has a critical chance of 8.5% for Zeal.