Squad Builder Online Resource

Hello everybody!

As I was playing squad builder yesterday, I was thinking about how nice it would be to know the best possible move to get the most points. So, I decided to build an algorithm that would take your current hand and recommend the optimal next move. You can then tell it what characters you got next, and it will guide you through all 3 rounds.

This tool is available online at:

https://squadbuilder.mysticartsarchive.com

I hope it is helpful to you and the rest of the community! If you encounter any problems, let me know.

Comments

JESUSCHRIST

Can you explain the logic behind the recommendation bcos I tried some examples and I don't agree with the recommendation provided

JESUSCHRIST

It seems that the advice provided is leaning towards being conservative

Doctor_Strange19

The model does tend to take a conservative approach. It assumes you will be playing a lot of rounds, and tries to get you the best possible score over the course of those rounds. If you want to play it riskier, you can tell the model what cards you chose instead of it’s recommendation.

Doctor_Strange19

Here is a more in-depth write up about the model’s logic:

When you enter your hand, the optimizer generates every possible way you could swap cards — discard 1, discard 2, all the way up to discarding all 5. For each of those options, it calculates the exact probability of every draw you could get back from the remaining pool. If there are 19 Fire and 13 Earth left in the pool and you draw 2, there's a specific probability for drawing 2 Fire, 1 Fire and 1 Earth, 2 Earth, and so on. It uses the known composition of the pool to compute these probabilities exactly, then multiplies each possible resulting hand's score by its probability and adds them up. That sum is the expected value — the average score that swap would produce over many games.

The optimizer also looks ahead. On Round 1, it doesn't just evaluate what your hand looks like after one swap. For each possible hand you could end up with after the Round 1 draw, it runs the same calculation again for Round 2, and then again for Round 3. So it's tracing every branching path: every possible discard choice, every possible draw result, across all remaining rounds. It picks the discard that leads to the highest average final score when you also play optimally in future rounds.

For example, say you have 3 Fire, 1 Earth, 1 Metal on Round 1. That scores 108 points as-is. The optimizer evaluates discarding the Earth and Metal (keeping 3 Fire, drawing 2 — with a chance of reaching 4 or 5 Fire over the remaining rounds), discarding just the Earth, discarding everything, and every other combination. Each option gets an expected value number. If "discard 1 Earth and 1 Metal" comes out at 542 and "keep hand" is 108, the 542 option is recommended because it produces a higher average outcome.

The expected value shown for each strategy is that average — not a guaranteed result. Sometimes you'll draw well and score higher, sometimes you'll draw poorly and score lower. The number just represents the average across all possible outcomes, weighted by how likely each one is.

SummonerNR

Nice, but some of the “calculates ALL possibilities” seems in excess and shouldn’t even be programmed into it.

Such as having 3 of a kind already. No need for having it factor in possibilities of not swapping any cards, or even just swapping 1 card (why would you swap only 1 of the 2 “solo” cards when trying to upgrade 3 of a kind into 4 or 5 of a kind, from your example).

And with such existing 3 of a kinds, probably more curious to know what the Odds would be of that turning into either 4 or 5, since the “what should I do then” is probably already clear.

(2 Pair scenarios…)
As for choice determinations, really would just like to know the Points/Odds (ie, Cost/Benefit ratio) between building on one color vs another (most importantly say if you've got 2 pair, which color is the best pair to keep statistically). Especially since Kabam structure the points around nice whole 8888 numbers, that may not truly reflect the Odds for sets of each of the different Colors. Was originally thinking of tagging in @DNA3000 for that (guess I just did).

As well as maybe an OVERALL POINTS AVERAGE per match, to know about how many matches it should take (mathematically on average) to reach highest Solo Milestone. (As a base # of points per match, and then also factor in the # of Free 3x tokens you get along the way to come up with the “if I do this all for Free” estimate of matches).

captain_rogers

Wow some people are so intelligent. I couldn't even imagine doing something like this lol😭.

peixemacaco

https://forums.playcontestofchampions.com/en/discussion/411328/squad-builder-online-resource

Thanks for this tool, but this isn't for Lucky people.

Of course Earth and Fire are both that comes again and again…

But even with only one Wood, I will try for more… Same for Water and Metal.

This tool is ok to prioritize medium points to reach milestones, but me and a lot of others will wait for a better luck and major points.

SummonerNR

My average so far.. (and estimates + info)

(* and side note for anyone else whose milestone points seem to be earning MORE that the points given for their Card Hands, exactly 32 points more each match, and no matter if using 1x energy or 3x tokens. Yes, seems to give 8 additional points per fight, so 32 points per match unless you freeze your hand early. Don’t recall seeing that mentioned in the announcements for how to earn points.)

14 Matches so far. Average (at 3x) of 1870 pts per match.

Which after subtracting the bonus 32 pts, and scaling down to 1x base value came to a 613 Avg. Which is right inbetween the points listings for 3-matching Blue and 4-matching Yellow hands (on the points table).

There are 67 total FREE 3x Tokens avail in milestones. Which would come to 125,290 pts (from my Avg above). With Top Milestone at 136,888. So would likely need to do a handful of Energy-cost 1x entries (if nobody sends you any of their unused 3x tokens as gifts). At a total of about 10 hands every day over the course of the week. (multiplied by 3 weeks duration in all, and your Rank each week potentially earning you weekly points into the Ally (3-week) cumulative Stampede event milestones).

CheeseMaster482

Good to know me and the algorithm think the same.