Hello,
We wanted to normalize skill builds to a specific level so that we can aggregate that data and provide a clear roadmap for people to use. For the purposes of aggregation, skill builds that don't reach level 18 are thrown out and those that exceed level 18 are truncated.
You make very good points. Let me ask: do you have an alternate approach that would ensure consistency of the data that's displayed and be more informative? We're open to making changes to the system if they'll be more helpful to players.
Give standard build progression to 11 and add the most common and most successful paths from 12 to 18 if applicable. Might work, might not, but it could be worth taking a look at.
I'm not sure if I can offer anything more useful, as statistics is not really my field, but I'll try.
Consider the highest level attained by a hero in at least 50% of valid games. The "median maximum level"?
Calculate your base build rate and win rate there, and then any mark the most popular progression of games where the hero reaches 18 as the final build. Note on the page the hero's median maximum level and some form of average level for that build at end game?
An issue with this is that there is a case where there are two final builds based off the same base build that could make the popularity cut.
The value is skill builds based on more games with win rates and build rates based on a level players can expect to achieve themselves without loss of the final build suggestions.
I'm sure you all have put a lot more thought into this than I have, but I hope some of my ideas help spark your own, fortified by greater knowledge and familiarity with the subject.
You should be able to put together a simple role classifier based on participation/contribution and from there, weight each build prefix by it's win:loss advantage. For popularity, prefixes contribute to all builds consistent to the prefix. At that point, display the most popular build of the most popular role with annotations of significant build divergence points for secondary and tertiary builds (for a lot of them, that I'm aware of, those divergence points will be on the first 2-3 levels) as well as annotations of significant advantage shifts.
Pro psaní komentáře se přihlaste.
For many heroes in many games reaching level 18 is highly unlikely. The skill builds shown on the hero profile don't seem to take into account the games that end long before a support hero reaches level 18. Not every game is 40 minutes or longer. Game 1 of TI2 finals not a single hero reached level 18 in 28 minutes.
If the skill builds are not counting the many games that do not last that long and only counts the stomps where one team totally trashes the other then its pretty uninformative. Even Meepo has a 65% win rate on his skill builds. Though winning with Meepo is probably a test of micro skill rather than minor skill build differences, reaching level 18 at all signifies the game is probably going well. It tells us that the team that has more exp wins 70% of the time. Shocking.
Can we get a clarification on just how the stats are compiled? How are shorter games or supports that may still only be level 14 at 35 minutes accounted for? Is there some sort of lumping that places unfinished builds with its most likely finished progression?