Gauntlet Calculations
calsnowskier
✭✭✭✭✭
in The Bridge
There are obviously some “hidden” bonuses in the gauntlet. Do we know any of them for certain?
Some that I have identified, but don’t have any actual multipliers...
“Defender Bonus” - The defender has a base bonus of maybe 10-15%
“Underdog Bonus” - Whomever is the underdog, whether the attacker or defender, seems to get about a 10-15% base bonus.
“Defender Crit Bonus” - The defender appears to have about an additional 10% added to their Crit percentage. If their displayed percentage is 25%, their true % is 35%.
“Attacker Crit Penalty” - The attacker seems to have their Crit percentage halved.
“x3 Streak Penalty” - The attacker seems to have about a 10% penalty to their roles when trying for an x3 streak.
I am not trying to be snarky. I truly believe these modifiers exist. I actually use the Underdog Bonus to my advantage. Has anyone else noticed other potential modifiers that exist?
Some that I have identified, but don’t have any actual multipliers...
“Defender Bonus” - The defender has a base bonus of maybe 10-15%
“Underdog Bonus” - Whomever is the underdog, whether the attacker or defender, seems to get about a 10-15% base bonus.
“Defender Crit Bonus” - The defender appears to have about an additional 10% added to their Crit percentage. If their displayed percentage is 25%, their true % is 35%.
“Attacker Crit Penalty” - The attacker seems to have their Crit percentage halved.
“x3 Streak Penalty” - The attacker seems to have about a 10% penalty to their roles when trying for an x3 streak.
I am not trying to be snarky. I truly believe these modifiers exist. I actually use the Underdog Bonus to my advantage. Has anyone else noticed other potential modifiers that exist?
0
Comments
I cannot speak for the others, but I did record crits vs. noncrits for both attacker and defender for a month and the displayed percentages are honest and I can't think of any reason the others wouldn't be.
You can believe in these hidden bonuses if you want to, but unless you actually have data it'll be hard convincing people.
Proving a great point that is very true with the forum group. The research is done. The homework is complete. The facts have been tested...
Remember, if you play matches that have an 80% chance of success, the odds of 12 consecutive wins is only 6.87%.
Even at 90% chance of success, the odds of 12 in a row are only 28.24%.
Gauntlet is designed to make it unlikely for you to get long streaks just by the nature of RNG. There is no conspiracy needed.
Almost 20k rounds on the gauntlet and while I can't say about all of that you listed I can say for sure if the "defender" has 25% crit chance and only 3 attacks is about 70-80% certain it will crit at least once plus any other "natural crit".
People here gathered the data, yadda, yadda, but it never fails. 3 attacks, 25% crit, and the game will score a crit. In a game like this, I really don't trust any of the displayed %.
Psst, the gaming industry don't care. Don't believe me? Check EA, Bethesda, Activision, Take Two, etc, etc...
If you have a 25% crit rate and 3 attacks, there is a 58% chance to crit at least once. The difference between that and your 70-80% is 1 one fewer crits out of 10. Unless you tracked your matches, there is no way you intuitively observed the difference between the two. So sounds like it's happening as often as odds say it should.
"5% player" = 43 crits / 834 rolls = 5.10%
"5% opponet" = 50/1227 = 4.07%
"25% player" = 451/1788 = 25.22%
"25% opponet" = 297/1236 = 24.03%
"45% player" = 344/555 = 43.96%
"45% opponet" = 293/642 = 45.64%
"65% player" = 113/171 = 66.08%
"65% opponet" = 62/102 = 60.78%
I would have liked to get more data points for the 65% tier. Aside from that there is about a 1% variation from displayed percentages. Cetainly not 10% and not favoring defender.
Phenomenal stuff @Odo Marmarosa !! It’s understandable you didn’t get a ton of 65% ones. That’s tough to get. And there you have it folks. The research has been done!
I bet there would be some interesting research available right on this forum for someone interested in both statistics and sociology.
I hope you don't feel people are going out of their way to disagree with you, it's just that this forum has a long history of in-game math theories followed by significant data gathering to prove the outcome. So far I'm only aware of one proved data point with large data collection, which was the AND shuttle display% bug. This bug seemed to be oddly stealth fixed without an announcement, around the time some other shuttle in-game changes were made to display.
I have a strong recollection that someone did some follow-up data investigation to prove the fix was applied, but I've been unable to find those old forum posts in a while so it's been an ongoing forum discussion ever since.
I don’t recall the post-fix evaluation to be anywhere near as rigorous as the work done to prove the bug’s existence but I think it was enough to prove the fix was made and works.
Last event really broke the dam for me on that since Navaar had no place to naturally fit. I put her in a MED and COM slot in a three seat shuttle all event long. I may have failed one of those shuttles.
Over the course of 4 days that’s a pretty good sample size. Something just keeps telling me not to do it
I still try to avoid it where I can, but in extreme cases like that I just go with it now.
I know what you mean, though to be honest I have gotten lazier about keeping to the old AND rule since I don’t feel strapped for most of the faction-only items any more. Example: last Friday night I realized I hadn’t leveled Amelia Earhart and that her ENG/COM would be high enough to be useful on shuttles. I power-leveled her all the way to 100 in about five or ten minutes, only needing to farm components I think for her 5* badge.
That’s how it goes with all event 4* crew as well...in fact, I don’t really remember the last time I had trouble equipping anyone. My first Yarnek took three weeks to get all the components together (trying to do Cyrus Redblock at the same time was probably dumb) but leveling my second copy was a cakewalk.
Results, over a large sample size, are exactly as they should be
Second Star to the Right - Join Today!
There’s a name we haven’t heard in a while
We have "all this data" and yet not a single one of these data collection efforts ever led to detection of the frozen gauntleteer bug. Theoretically this bug has existed since the origins of gauntlet. Nobody but DB will ever know the truth, because this data is inaccessible.
Moral of the story: if you don't know the conditions you're testing for, an aggregate is virtually meaningless.
IAmPicard meanwhile presents an interesting intersection of statistics and sociology. Namely, what happens when a portion of the player population readily has access to more and better information than the rest of the population? It couldn't possibly influence outcomes could it!? Anyone have a peer reviewed study to drop on us common folk???
Coverage bias is pretty basic, and yet there have been zero attempts to address it.
(Let's not forget the whole cheating scandal. Surely we can trust IAmPicard derived data points, because "a lot" of it was collected. The rehabilitation of IAmPicard begins here.
Regarding non-IAmPicard data sets... manual collection of large data sets doesn't sound the least bit problematic, does it?
I work with people for whom "attention to detail" is one of the top three professional responsibilities (arguably the top responsibility) and I was horrified to learn how rare attention to detail actually is. It's probably out there past shark attack odds. I could provide numbers, but this is proprietary information.
At best, the data made available to forum-goers points to the possibility that match outcomes basically align to expected results in situational-agnostic conditions across what amounts to around a year of a passionate gauntlet player's matches. Cool beans. But that's not the game 99.9% of us are playing and that's not what the OP is discussing.
(Shout out to all the confirmation bias in the house. Put your hands in the air now! 🙋)
Let's not forget the third rail of gauntlet stats debate: match selection data, or more accurately, the glaring lack of it. If you don't implicitly understand how this information is likely key to deciphering situational match outcome stats, I don't think there's much I can do for you.
This is one case where I wish IAmPicard would have used their developer powers for good rather than evil. But they didn't. We might as well call this data inaccessible since it's not only about the matches you see, but the larger pool of potential matches out there.
So when even the top tier data collection effort ever presented here has less than half of the necessary data needed to determine conditions, suffice to say, I ain't boarding anything going to space on data like that. I'd be dubious about using a scooter built on data that trustworthy.
Sure, observational bias is a thing. And yet, shock of all shocks, people successfully diagnosed problems before we had robust modern data science. 🤔🤔🤔🤔🤔
(They were undoubtedly witches and should be tarred and feathered as such in public spaces.)
Now then, please continue with the size comparison of your data sets. I find it fascinating.
I believe that statistics work fine in Gauntlet but due to the subjective perception of outcomes it fosters suspicion. The implementation is just frustrating per design which is poor for a game. It should be fun instead.
The reward structure and time zoning does the rest to totally frustrate players. It is a wonder that still so many people even care to take part without it being relevant for the daily goals. I am one of them but I increasingly dislike it.
Match selection may be important in the overall gauntlet, but that's not what people complain about. The complaints are basically, "I lose too many matches I think I'm supposed to win". The data collected so far says the individual matches are fair.
Yes I wish the data was better, yes I wish there was more we could look at. But to look at what was collected, change the question being asked, claim there is no data for that question, therefore you support the original conspiracy theory? That's not exactly data science either.
You’re absolutely right that there are still many questions to be asked about the gauntlet. Match selection bias was something I asked Rex to look into, because my contention was (and still is) that there is a higher likelihood of getting a skill match from a character who just lost (thus influencing the player to want to revive the recently-defeated character in order to match those skills properly), but he wasn’t able to use the data collected to analyze for that. As much as confirmation bias is probably at play, I cannot shake the idea from my head that the gauntlet is tilted in this manner.
You're talking abut a couple of separate things there. Why would there be a need to intentionally stack opponent choice against you to drive more dil spent, when the sheer volume of gauntlet rounds played will inevitably mean that you've got poor matchups regularly via RNG? It's extra effort programming in some conspiratorial small potential revenue gain, which is in my opinion irrationally unnecessary.
Your other comment about crits being sometimes lower values than opponent crits, you can see that in your character stats. There is a RANGE of stats which is used as your "roll of the dice". Let's say it's 100-200. A crit from a 100 roll is significantly different outcome from a 200 roll crit, and the ranges are typically much larger than that from minimum to maximum. There are 2 layers of RNG here, for the initial dice roll then the RNG giving you a crit chance. If your crew has a larger skill roll range (I think phlox is an example of someone with a large range), you are more susceptible to RNG than someone with a smaller range of possible rolls for the same skill.
I'm glad this game is made up of Star Trek fans. Can you imagine how much complaining about math would happen from fans of a drama show? I'm surprised at the volume of complaining about math happens on this forum for fans of a science fiction show, whom I'd anticipate being a more analytical audience than average.
Obviously.
This arguement makes no sense to me. Of course it would make sense to trap people into spending dil by giving them opponents they could beat with dil-refreshed crew but not with what they currently have. But I did not even say that it is that way. I just said that the choice of opponents is intransparent. Or did you find a pattern so far? Like "1st is 20 ranks higher, 2nd is 10 ranks higher, 3rd is 1 rank higher, 4th is 10 ranks lower and the last is 20 ranks lower" (given that you are somewhere in the middle and not in the top or bottom 20? You could not even if you tried as you only can see one or two small parts of the ranking at a time.
Many people complain how hard it is to get a Locutus. Still I am confronted with pure Locutus walls incredibly often. How likely is that if the opponent choice is completely random or follows a defined pattern like in my example above?
I know. But my biased (!) impression is that the opponents more rarely have such annoying bottom range crits but rather extraordinary one. Of course this is easiest to observe with identical crew and stats as no calculations are required.
This is a bias too.
I am merely complaining about the design and reward structure of the Gauntlet and not the math. It is just not enjoyable and frustrating. You only need to look at the result animation which most of the time shows the end result on the halfway to understand how uninspired they were when they implemented Gauntlet. And I am not convinced that it is designed and programmed in a way that fulfills the criteria of randomness I would expect from a proper implementation. (Of course, I worked in the Casino gaming industry and my expectations on randomness are accordingly high.) At some points they probably had several options how to proceed without having to give any explanation to the community (e.g. opponent choice) and perhaps chose the one that promises the most profit? Like hard-but-with-dil-possible-to-come-by opponents?