I applaud your efforts and look forward to hearing about results. The hardest thing to tease out is player skill (maybe honesty too).
It seems to be the more niche heroes end up with the best win rates. Are those the best heroes? Are they only used by players who set them up well for a win?
Yeah, getting rid of the outliers of heroes that only a few people use will be an issue. I am hoping to partially get round that by excluding any hero I dont have at least 100 records for. But for borderline heroes the results will be artificially inflated. Maybe I will include percentage of error also to reflect that.
Player skill/ honesty will always be an issue to get right I am hoping if can get a large enough data set it might limit its effect though.
The idea for this came to me in the shower while thinking about some payment plan breakage analysis I am doing for work currently along with the usual this hero is trash/op threads. Its be nice to see some kind of details to look back to
The big thing that is not being factored in (as far as I can tell) is the opposition. The TP, the tank, the other heroes, the raid formation…
Someone may come to you with 90/100 cherry-picked wins for a weaker hero whilst another person will have 50/100 of a stronger hero but facing off against much stronger opposition
Adding in all the defence variables is currently too much sprawl and while useful is limited as TP by itself is meaningless as opposed to actual team composition. The limiter I have in it to deal with cherry picking weak opponents is raid arena level of the defence being faced. Meaning that can at least say hero A has a good winning percentage in an arena level overall to provide a guide as to potential usefulness
I get the difficulty of introducing all those variables. But there is a long long way from the easiest diamond defenses to those at the top.
Perhaps stipulate that anyone that provides data either cannot reroll or else they specify an exact defensive criteria (eg summoner tank) which will provide an insight on the versatility of a hero vs nichenesd and potentially cut down on cherry picked outcomes
Aspects of that may be useful. I will try to have a think as to how to both incorporate that with this whilst still being easily accessible for information or whether to make a seperate file with details versus types of tank. Which would be an additional later step after this.
If there are enough number of volunteers willing to spend some time entering the data, it’s quite straightforward to create a Google Sheet for entry like this
So I’ve made a template now. To make it as easy as possible for people to complete I’ve tried to simplify the fields required and to address some of the points raised by @Homaclese I’ve added a column for tank faced so the final data will be filter able by that too.
Anyone wishing to help I will forward the template to and they can then send me sheets by either line or email to contact details in first post and I will collate and produce a finished version.
No offense, but this is just not really achievable…waaaaayyy too many variables to make this meaningful IMO. SGs list of “best 5 defense heroes” and “best 5 offense heroes” is about as good as we can get. Have they released one of those lately?
None taken. I am not underestimating the size of what this entails and all of the defence formations etc was one of the reasons I dont want to get into the kind of project sprawl that would involve.
The main thing I am looking to do with this is show hero usefulness etc as I see so many blanket statements about needing certain hero types etc to win raids at different levels.
But a lot will come down to being able to collect enough data points.
