Snapalytics: How to Play Even More Like the Best

A new season marks the perfect opportunity to analyze high-level game stats. LaurenWhatever guides you through the numbers and highlights data-based insights that might help you win more cubes!

A couple weeks ago, I used data from our Marvel Snap Tracker to examine how cube rate (the average number of cubes won per game) is impacted by an assortment of different metrics (measurements). I highlighted the best of them with charts and numbers.

It may be counterintuitive, but when you want to find the most competitive stats for ranked mode, Infinite tier (rank 100+) is the worst place to look. The fact that players who reach Infinite can’t drop below rank 100 for the rest of the season means that most of them play a lot more casually once they get there. They usually are still trying to win, but they run less competitive decks and rarely (or never) retreat.

Because of that, it’s best to separate or exclude Infinite-rank matches when analyzing game data. I excluded Infinite matches from my analysis a couple weeks ago. CalmingSkylark commented on that article and pointed out a problem with running the analysis so late into the season — the players who are best at climbing ranks weren’t represented in the data, because they had already reached Infinite.

Well, now we’re almost two weeks deep into a new season. It’s time for a sequel.

Bigger and Better

This time, I’m covering a data set that’s roughly twice the size. That’s mostly due to expanding the qualifications for the sample group. I tested different groupings based on ranks and collection level. The groups had very similar data, so I’ve combined them. There were a few differences worth mentioning; I’ll call those out as I get to them.

Data in this article covers matches that meet the following requirements:

  • In the first 11 ½ days of the newest season (Guardians Greatest Hits)
  • Player rank was 30–99
  • Player deck did not contain Agatha Harkness

After that, any player who didn’t record at least 60 qualifying matches was removed from the data.

Agatha isn’t useful for studying player decisions because she takes so many of them away.

The start of a season is an excellent time for this analysis, because ranks get reset. Everybody who was in Infinite last season started this season at rank 70. That means the data includes our best rank-climbers. About 8% of the players in the data set had already hit Infinite!

Our sample size is up! I investigated more metrics!

Players: ~2,000 ↗️ ~4,800
Matches: ~550,000 ↗️ ~1,010,000
Metrics: 42 ↗️ 52

On “r”

Each metric will come with a chart and I’ll talk about the data. One of the most useful values is the linear correlation, which is represented by the letter r. Linear correlation looks at the straight line that best describes the data, and then considers how closely the data actually fits that line. Practically, the correlation tells us how related two measurements are. The main measure we’re interested in is cube rate.

r is always between -1 and 1. The number represents both the strength of the relationship and the direction of the relationship.

r = -1: Perfect linear correlation, negative relationship (metric ⬆️ = cube rate ⬇️)
r = 0: No linear correlation at all
r = 1: Perfect linear correlation, positive relationship (metric ⬆️ = cube rate ⬆️)

Most metrics we’ll look at are weakly correlated. Our sample size is large enough to ensure the correlation is not due to random chance. A weak correlation doesn’t mean a metric isn’t relevant or important. Instead, it means that the metric is just one small piece of the puzzle that is cube rate.

Remember that even small improvements in your cube rate add up over time. It takes 168 cubes to climb from rank 70 to 100 (counting the bonus 30 you’ll get when you hit rank 80 and then again at rank 90). If your cube rate is 0.1, that’ll take you 1,680 games. If you improve your cube rate to 0.105, it’ll take you 1,600 games (saving 80!).


Metric: Win Rate

Unsurprisingly, win rate has a strong correlation with cube rate. After all, you only gain cubes when you win.

r = 0.76 (strong and positive)

For every 1% more of your games that you win, your cube rate is expected to increase by 0.046. Of course, winning isn’t everything. The snap and retreat mechanics make it more complex than that. Just look at how many players on the chart have positive cube rates with a win rate below 50% (or even below 40%!).

Metric: Win Rate when Nobody Snapped 🆕

Winning is always important, but it’s a lot less important when the stakes are low due to neither player snapping.

The data in this section requires that players recorded at least 20 games where they didn’t snap.

r = 0.25 (weak and positive)

For every 1% more of your games where nobody snapped that you win, your cube rate is expected to increase by 0.009 cubes per game.

Snapping & Retreating

Snapping and retreating determine how many cubes are at stake, and so they are absolutely critical to rank climbing.

My First Infinite: Climbing Mindset

Last season was the fifth I played but the first that I hit Infinite. I attribute that to changing the way I think about the game. I shifted my focus from winning each match to winning across an entire session of play.

Over many matches, the real game of Marvel Snap is reading the board state correctly. How likely are you to win? Your primary goal becomes the correct identification of when you should snap or retreat. Winning is a secondary goal. In my case, that meant playing a lot more conservatively — I’m naturally impulsive. It also meant finding a deck I liked and sticking with it. Developing familiarity with my deck helped me snap and retreat more accurately.

Metric: Snap Rate

Overall, snapping more often improves your cube rate. That makes sense: if you never snap, you’ll be playing for 4 cubes less often and will never be playing for 8 cubes.

r = 0.04 (very weak and positive)

For every 1% more of your games that you snap, your cube rate is expected to increase by 0.004 cubes per game.

Metric: Average Snap Turn 🆕

It can be useful to know which metrics don’t matter. This seems to be one of them. r is miniscule and fails to be statistically significant (88.20% chance the correlation is due to random chance). On top of that, the trendline is nearly flat.

The data in this section requires that players recorded at least 20 games where they snapped.

I just thought this related chart was interesting. It’ll be the only one in this article where the vertical axis isn’t cube rate. I’m fascinated by the cluster of players that snap on turn 1 almost every game!

Metric: Percent of Snaps on Specific Turns

Here is one area where the groups of players diverged based on rank. There isn’t a single snapping turn with a correlation worth mentioning among players ranked 30–59.

At ranks 60–99, there are three snapping turns worth mentioning: 1, 3, and 4. Interestingly, when I was looking at data two weeks ago, it was turns 3, 4, and 6 that were worth mentioning. That shift could be representative of a shifted metagame, or perhaps it’s the difference introduced by including the best rank-climbers before they hit infinite.

The data and chart in this section will cover ranks 60–99 only. 

The data in this section requires that players recorded at least 20 games where they snapped.

Turn 1

r = -0.11 (weak and negative)

For every 1% more of your snaps that you do on turn 1, your cube rate is expected to decrease by 0.005 cubes per game.

Turn 3

r = 0.07 (very weak and positive)

For every 1% more of your snaps that you do on turn 3, your cube rate is expected to increase by 0.003 cubes per game.

Turn 4

r = 0.07 (very weak and positive)

For every 1% more of your snaps that you do on turn 4, your cube rate is expected to increase by 0.003 cubes per game.

Metric: Percent of Snaps You Retreat

This metric and the next are about snap accuracy. Are you snapping when you shouldn’t have? You shouldn’t be too proud to retreat after snapping if you find a match suddenly favoring your opponent — but even better if you hadn’t snapped in the first place!

The data in this section requires that players recorded at least 20 games where they snapped.

r = -0.17 (weak and negative)

For every 1% more of your snaps that you end up retreating, your cube rate is expected to decrease by 0.013 cubes per game.

Metric: Percent of Snaps You Win 🆕

More snap accuracy. The better you snap accuracy, the more often you’re winning when you snap. Note how the correlation on this one is quite high, like the other metrics based directly on winning.

The data in this section requires that players recorded at least 20 games where they snapped.

r = 0.46 (moderate and positive)

For every 1% more of your snaps that you win, your cube rate is expected to increase by 0.019 cubes per game.

Metric: Retreat Rate

Retreating is complicated. If you’re going to lose, you should retreat, which can halve or even quarter your cube losses. However, retreating counts as a loss, and we’ve seen how important win rate is to cube rate. Retreating minimizes losses, but it’s still a loss.

r = -0.20 (weak and negative)

For every 1% more of your games that you retreat, your cube rate is expected to decrease by 0.009 cubes per game.

Metric: Percent of Losses You Retreat 🆕

Here’s a metric that better represents the value of retreating. Like I said, if you’re going to lose, it’s better for you to retreat.

The data in this section requires that players recorded at least 20 games where they retreated.

r = 0.18 (weak and positive)

For every 1% more of your losses that are due to retreat, your cube rate is expected to increase by 0.004 cubes per game.

Metric: Unique Decks Played

The more experience you have with a deck, the better you’ll become at correctly reading the board state — to accurately guess whether you are in a position to win or lose. That’s why I’m putting metrics about deck usage in the section on snapping and retreating. Deck mastery should improve your usage of the snap and retreat.

r = -0.26 (weak and negative)

For every additional deck you play, your cube rate is expected to drop by 0.015 cubes per game.

Note: There is more to this metric than just a lower number of games making it easier to get extreme cube values. The correlation and effect remain similar if you only include players who recorded 200+ games.

Metric: Percent of Matches Using a Deck Already Used 20+ Times 🆕

I admit, this one is awkwardly named. It’s another metric about deck mastery.

This is another metric where the results markedly differ by rank. It’s statistically significant for both groups, but the correlation and effect are both roughly twice as strong for players in ranks 60–99.

The data and chart in this section will cover ranks 60–99 only.

r = 0.15 (weak and positive)

For every additional 1% of your games that you use a deck you’ve already used 20+ times, your cube rate is expected to increase by 0.002 cubes per game.

Note: The players who have very high percentages (including 100%) reached their 20+ “training” matches before they reached rank 60.

Take Your Time

Metric: Average Turn Seconds 🆕

How long do you sit and think before hitting End Turn? Personally, I find myself saying “Oh, sh—!” right after submitting my turn more often than I’d care to admit (Undo button when, Second Dinner? 🙏).

This metric matters quite a bit in ranks 60–99, and almost not at all in lower ranks. I’ve tried slowing down (especially on turns 5 and 6), but this is still a big area of opportunity for me.

The data and chart in this section will cover ranks 60–99 only.

r = 0.14 (weak and positive)

For every additional second per turn that you spend (on average), your cube rate is expected to increase by 0.017 cubes per game.

Just for Fun

Metric: Matches Played

I just thought this chart was cool. It shows that it’s hard to maintain a very high or very low cube rate if you play a lot of matches — nobody who has played over 300 matches has a cube rate over 1.0.

I think the right end gets skewed by players who are farming season pass XP and/or running Agatha. The big empty space at the beginning is because this data set is only looking at players with 60+ games.


Remember that all this data is covering trends. Some players and decks can buck the trends and still perform very well. Just look at some of the outliers on the charts above! There are players with high cube rates that almost never snap and others that almost always snap. Some have positive cube rates despite losing over 60% of the matches they play!

The high variety in ways to be successful means weaker correlation and less useful data for shaping your gameplay, but I also love that there are so many ways to play that can viably rake in the cubes. If I were to sum up the takeaways, though, this would be my general advice for improving your success at climbing the ladder:

  • You don’t need a top tier deck, but you do need a deck that’s strong enough to win roughly half of the games you play
  • The goal is not to win every single game, but to improve your cube rate across many games
  • Take the time to master the decks you play (especially to improve your snap and retreat accuracy)
  • Snap more often when you’re going to win!
  • Retreat more often when you’re going to lose (especially consider retreating on the turn your opponent snaps)!
  • Slow down

Climbing the ladder is about cubes, not wins. That seems obvious, but I think too many players don’t go far enough to change their strategy accordingly. Or maybe that was just my own experience. 😅

The data-based insights throughout this article provide valuable guidance for increasing your cube income. Perhaps the most important area to improve is the accuracy of your snaps and retreats. Part of that comes from mastering your own playstyle and favorite decks. Part of that comes from staying knowledgeable about the current metagame and adapting accordingly (something Marvel Snap Zone articles can really help with!). And hey, it’s not all about hitting Infinite. Each season leading up to my first Infinite, I climbed higher than I had the season before — and that always felt good. Reaching 40 is amazing if you’ve never done it before, and you should be proud! Good luck on the rest of your climb!

This analysis was powered by the data from our Marvel Snap Tracker. If you play on PC or Android, consider giving it a download! It has many features including a live overlay, the ability to sync your collection to our site, letting you review your personal stats, and powering useful tools like the Meta Snapshot.

Enjoy our content? You can Support Marvel Snap Zone and your favorite content creators by subscribing to our Premium community! Get the most of your Marvel Snap experience with the following perks for paid membership:

  • No ads: Browse the entire website ad-free, both display and video.
  • Meta Reports: Exclusive daily meta reports, such as the Top 10 Decks of the Day and Trending 30 Cards tailored for you!
  • Early Access Tier List: Exclusive early access to the weekly Tier List update (decklists, rankings, and article) 24 hours before the article goes live! Get an early start on your competition!
  • Premium Dashboard: Get full instant access to the member-only dashboard, the all-in-one page for all your benefits.
  • Support: All your contributions get directly reinvested into the website to increase your viewing experience! You get also get a Premium badge and border on your profile.
  • Discord: Join our Discord server, claim your Premium role and gain access to exclusive channels where you can learn and discuss in real time!
  • Special offerFor a limited time, use coupon code SBYREX4RL1 to get 50% off the Annual plan!

Lauren likes games. Mostly for fun, but she has competed in TCGs and Smash Bros. She is here to crunch numbers for her beloved Marvel Snap after (accidentally) spending 7 years as a data analyst. She lives in Utah with her fiancée, 11yo, and 2 very good dogs.

Articles: 9


  1. Hi Lauren,

    I think you should put the conclusions/insights at the top of the article, like a tl;dr. Everyone that clicks on this article would see the takeaways right away, and the people that would want to know ‘why’ (like me) would stay for the meat and bones.

  2. Hey Lauren, great job as always. Your articles are quickly becoming one of my favorite parts of snapzone. Just wondering if it is possible to measure the thing that you did last season: first time to infinite. How many players that hit infinite in a season are hitting it for the first time?

    • Ooh, that’s a fun idea. It’d be a little tricky because it depends on players using the tracker over multiple seasons — but I’ll give it a try!

  3. Lauren,

    This report is amazing! My biggest takeaway is that your “Matches Played” metric makes the game look like a vegas casino, “the house gets its money back over time.” If it’s the case that nearly every player regresses to the mean over time, then the game is too random.

    Is that how you see it?
    Thanks again for the great insights!!!

    • Hm, partially. The game intentionally does matchmaking to put your win rate at about 50%. The snap and retreat mechanics are how you beat the house — to come out with positive cubes despite a 50% win rate.

      The mean for cube rate seems to be around 0.25, so it’s entirely still possibly to climb with a high match count.

      I also think farmers are influencing the data. How many players knock out 500 games in two weeks? I’d be interested to see what long-term cube rates look like over several months instead. A small advantage in cube rate goes a long way — if skill can keep you at 0.5 while most players sit below 0.25, that’s huge!

      That said, if we still see a severe regression to the mean then you’re right. Some randomness is good, but it’s TOO random if skill can’t let you beat the mean by a decent margin.

Leave a Reply