Many of you have asked us to share what we are working on and the math behind the suggestions we do. We will try to do that from now on.
Currently we're doing some server side improvements, trying to add tips for all the heroes but also searching for something to replace the time graph, which isn't very accurate.
In its current form is says, at every minute of the game, what are the chances to win or lose the game at that minute. This is not informative at all for the early game (few games finish before 20 minutes) and it also favours the heroes which push high ground easier.
For each game the Steam WebAPI offers, for each hero, the exact time at which an ability was learned. This allows the possibility to approximate the XP each hero (and each team) had at certain intervals.
This means that through some mathematical interpolation we can know at a each time in the game which team had an XP advantage and how big.
If we do this for hundreds of thousands of games we find out the probability that a hero brings an advantage to the team at a certain point in time. This should be a lot better than what was used before.
If this all goes ok and the results are promising we can go even further. Some heroes will give a better advantage in certain parts of the game (early, mid or late). We can thus statistically calculate some scores for this and add them to the suggested heroes on request. We haven't decided on the formula yet, only that the sum of the scores should add to 0.
We'll take an example to make this a bit clearer. Let's say it's our turn to pick and we notice (from the time advantage graph) that we have a team that is strong in the late game but kind of weak early. Let's say Undying has +1.2 in the early game, +0.5 in the mid game and -1.7 in the late game. In this case 1.2 would be added to the score of Undying if we would select we want bonuses for early game and 1.7 if we want bonuses for early and mid.
We're not sure if this will work, we already had some failed ideas (like trying to suggest builds statistically), but we will try to keep you up to date. Also the list of what we're working on will always be on the beginning of our changelog.
We have done most of the work and are almost ready to release it. The time graph is a lot better now. We also have 77% hero tips coverage.
The counter and synergy scores have been scaled to be between 10 and -10, for both normal and winrate modes, to give a unitary experience.
Bonus scores for early, mid or late game were added. These add or remove a percentage (up to 50%) of the full game score of a hero, depending on how good he is in that part of the game. An exception is made for the heroes which have full game score in absolute value lower than 1, so that the bonus can remain relevant- the bonus score is a percentage of 0.5.
We created a graph with all the data and added it to the Analyses menu.
All the contents of the update are live. Hope we improved your drafting experience.
We will now take some time to refactor our codebase so that we can add more features with ease. This addresses the requested improvement for UI on slow internet connections (1MB of data must be downloaded before the UI is generated).
Also we will soon ask you to help us shape the future of the picker so stay tuned.
As always, if you like what we do and can afford it you can donate to help us pay for hosting and improve the website.