Bureau of Rift Statistics · Corrections

What we got wrong, and what changed

Report No. 2

None yet. New errors go here, at the top, with their date.

Report No. 1

Every number in the report was checked against its source before it was published. The checks found errors, and this is all of them: what was wrong, what it is now, and whether any finding moved. New errors go at the top, with their date.

In the analysis

Date	What was wrong	What changed	Did a finding move?
23 Oct	Two movement abilities were missing from the kit table: Briar's R (she flies to the champion it hits) and Zaahen's R (he leaps to a target location). Found while checking the dash labels.	Both are added. Every escape-tool result was run again with them, on the same data and the same 2,000 resamples.	No. The largest change: in Iron, extra time dead per escape tool went from 7.3 to 7.7 seconds per 30 minutes.
16 Oct	The weights did not follow the locked plan.	Each tier is now weighted by its population over the accounts drawn, as the plan says.	Only the pooled share of time dead: 11.24%, was 11.33%.
16 Oct	The explanation of why the analysis has more rows than appearances was wrong.	Corrected: 973 games were drawn by two sampled players and appear once per draw.	No.
15 Oct	35 games from other Americas servers (LA1, LA2), from 7 transferred accounts, had been counted.	The whole run was repeated on NA1 only.	Every figure moved by 0.02 points or less. One unranked cell changed band: Zilean in Platinum, from below typical to far below typical.
15 Oct	The death-count model in the plan had not been built in the first run.	Built (negative binomial). It is published with its limit: it underestimates games with no deaths.	No.
15 Oct	The K'Sante test was written without checking how often he is played.	He can be judged only in Master to Challenger, and the report says so.	The claim is tested only where it can be.
15 Oct	"Noxus spends the most time dead" was mostly its mix of fighters.	The line was dropped once regions were compared within class.	Yes: Bilgewater, not Noxus, stands out.
before 15 Oct	The first keyword list of escape tools missed real dashes (it found 71% of movement tools).	Every kit was checked against the wiki's lists and by hand, before the data was opened.	No result had been computed yet.

In the article

Date	Was	Now
16 Oct	Patch 11.13 came "six weeks" after the design note.	About a month later.
16 Oct	Games at the top run about 27 minutes.	About 28 minutes (27.9).
16 Oct	Dash-heavy champions die "where nobody can save them".	Cut: it was not measured. What was measured stays: they die farther from their own towers, on the enemy's half.
16 Oct	Garen called the game's "marketing face".	Replaced, since no source supported it: Garen has been in the game since April 2010.
16 Oct	Garen "fourteen years" (twice).	Sixteen years.
16 Oct	Yuumi lowest "in every tier".	Lowest in 7 of 8 tiers. Milio is lowest at the top.
16 Oct	The Shadow Isles lead "in revival".	The Shadow Isles have the most untargetability. The kit table has no Shadow Isles revives.
16 Oct	A death-recap image from a designer's portfolio.	Removed: the designer couldn't be identified or asked.
16 Oct	The article's foot carried only the developer-policy notice.	Riot's Legal Jibber Jabber line is added wherever Riot's art appears.

Found something else? Tell us in the common room. Corrections are listed here with their date.