Solved: How does Pix determine Ag payouts (4/2019)

For BR wins it will be a bit higher than 0.2 because of the kills/beacons you get plus the 75k intercept on the graph.  This is the non-premium number.  

When I back checked my model to my data I came with 10% of actual 90% of the time.  There were some outliers and at least some of that has to do with critical hits not being counted.  pix says in their in game FAQs that what matters is damage, beacons, kills, and critical hits.

I can’t say I explicitly looked at topping damage or beacons.  When I play normally, I do tend to top damage and I am normally first or second on beacons.  However there were many games in the data where I was running around trying things out like only getting healing damage (didn’t fire at reds) or avoiding getting beacons to simplify things and all those games still fell on the trend.  I think it is safe to say that at moderate or high damage levels any bonus you might get for topping things, if such a thing exists, is lost in the noise.  If you are playing in silver league where damage is very low something like that might show up.

RepaIrs have nothing to do with actual damage taken at this point.  I ran several games with only a single bot from my hangar and got the same amount of repairs as running my full hangar so long as the damage I did to the reds was the same.  The biggest outstanding issue is whether or not old school bots somehow have lower repairs than newer bots.  I did try and test this a bit and didn’t see anything.  I’ve reached out to a forum member to get some more info on the subject which I will incorporate in the next update.

Good luck on your tests and let me know if you see something out of line.  I give my “error bars” for my back testing in the Word document so you can see how things compare.  If you get within 5-10% of what the formula predicts on a regular basis I’ll be happy.  There are things out there like critical hits that I couldn’t quantify and there also might be some sort of random thing on Pix’s end where the Ag they give you is +/- X% of the value they calculate for your payout.

I have had rounds this past week where I did not die during the match and my repair was like 10k, repair still counts. Bots and individual weapons are worth something like 30k. You don't pay for the weapons on a bot that died, just the ones that broke. 

It does seem odd that there is a high correlation between repair costs and damage done to reds. Seems counter-intuitive, which actually makes sense because so much of what Pix does IS counter-intuitive. But there are numerous examples of 0-low damage resulting in 0-low repair bills.  

Very interesting. Let me go through this to make sure I understand it correctly.

BR wins. In Pixos logic, silver before repairs is
Silver before repair = 0.403 * (Damage you do to reds) + 10000 * (# of kills you get) + 2400 * (# of beacons you grab) + 100000

Premium gives 50%, damage is substracted after the 50% gain, giving premium players almost a boost by factor 2.
Premium silver before repair = 0.604 * (Damage you do to reds) + 15000 * (# of kills you get) + 3600 * (# of beacons you grab) + 150000
Premium silver (minus repairs) = 0.398* (Damage you do to reds) + 10000 * (# of kills you get) + 2400 * (# of beacons you grab) + 125000


Adding a 40% silver booster if you run premium does what exactly? Do they add 90% on the base payout (50% premium + 40% booster) or 110% (1.5 times 1.4 for a 40% booster on top of a 50% one)?

The base payout has to drop below a certain damage level. If you start a new account and do ~100k damage, the payout is around 75k silver I think. Then, they wouldn't want to allow tankers to play for half a minute, get 50k damage in and then sit out the rest of the game, collecting 125k silver plus damage-based silver.

Wait... so the repair cost isn't for repairing OUR bots, but the bots we damaged?? That is weird...

Did you try a game where you just ran around with a single bot, did no damage and then let the bot be destroyed and then don't drop anymore? Then, see how much the repair cost is. This way, you are ruling out the damage you do, and you only have 1 bot from which to calculate the repair cost.

Yes... I know I'm suggesting that you troll a game... but heck, it's a random game where you don't owe anyone anything and it's all for data gathering so we all would benefit from it.

I did not test that scenario.  The closest I came was running a single bot in 5 separate games and once the bot died, I just watched the rest of the game.  I ran a Pursuer once, an Ares once, a Spectre once, and a Mender twice to see if bot/weapon selection led to significant differences.  Here is the graph of those games superimposed on my BR losses.  The five games are all at the low end of the damage range and I came in last or nearly last on damage in at least a couple of them.




Except for the lowest damage one, the points all fall nicely on the payout and repair trends.  The point that doesn't was my second Mender run that came right after the first Mender run.  Its possible that Pix does something to boost your Ag payout if you have 2 or more bad games in a row.  The graph and my description of things can be found on page 17 of the Word doc.

Anyway, thanks for the info.  I will do a few runs to test things out and will add that info to the next version of this.  If you have anymore of those runs, please send me the following info:

Beacon captures, kills, damage to reds, repair costs, Ag payout (non-premium), state of your bot at the end of the game (along with bot and weapons on it at start of the game).  I will then see if I can overlap your data with the data that I generate.

Repair cost used to be lower and was based on the damage taken by the bot you're in when the match ended. I used to switch bots with 5 seconds left in the match and had zero repair costs all the time. This changed at least a year ago.

Nice job!

Did you do this in a single bot that made it through the whole game?  That seems to be a problem with my model that I need to address.  Please send me data if you get high damage and loss repairs while losing at least one bot in the game.

Since repairs for the 1+ bot losses aren't really tied to damage you take but rather to damage you inflict, it works the same way as payouts.  In losses your repairs will be less for a given amount of damage inflicted on the reds than in wins.  That is what threw me at first.

Here are the two formulas for repairs in BR games:

BR win: Repair = 0.2056 * Damage done to reds + 5000 * # of kills + 1250 * # of beacons captured + 25000

BR loss: Repair = 0.130 * Damage + 5000 * # of kills + 2500 * beacons + 2250

The ratio of the slopes is almost exactly 1.5 (wins give you 1.5x the silver payout and 1.5x the repair costs ironically).  The other figures are slightly off on the 1.5x ratio but I assumed this was due to lumpiness in the data I have.

As to doing zero damage in a game (Kyle Rogers scenario), I played one game where I went around grabbing beacons and not firing on reds.  My thought was to try and get the "pure" value of beacons for my formula rather than having Excel calculate it.  That was a mistake.  I got four beacons, zero damage, 760k in repairs (not a negative number, a positive #), and a payout of -760k (negative not positive).  So I owed Pix 760k Ag for my efforts in that game.

I've listed the formulas for repairs, non-premium Ag, and premium Ag in the second post of this thread. I also posted a calculator in the same post so you can take your data and see how it compares.

If you consistently get results that are off by >10% or if you get something wildly off, please send me the details of the game so I can figure out what is happening. I will look into trying to make a single bot last through the whole game.

Very true. It seems that now switching  your robots will count as repair value.
I remember making a lot of silver without repair cost too.

Interesting.  Here is the in-game FAQs on Ag:




The FAQs doesn't mention time and damage taken and the data doesn't support either with the probable exception of somebody making it through an entire game with a single bot. 


I wonder what would happen if you ran a single bot hangar... would your repairs be 1/5 of normal since you have a single bot or would it match up with a full hangar.  Don't want to find out by spending 7500 Au but if anyone has a single bot hangar, let me know.

Thanks very much, linearblade!  I'll add them to the database and see if there is anything different going on.