5 replies to this topic

[Katotonic]

Hi all,

I will start with a bit of a non-sequitor. How many of you have watched the movie or read the book Moneyball? If you haven't I will VERY briefly summarize. They study all of the stats of individual baseball players to study their strengths and weakness allowing them to realize that as a TEAM they need a certain on base percentages, hit percentages, what type of hit percentages, etc. By reaching the TEAM target, they generally win. So you don't need a roster full of all-stars and a few weaker players to win, you can win with a balanced team that both enhances the teams strong points and covers for weaker points.

Now, how does this apply to MWO you might ask? Well, while it is a team game I am actually less interested in that than in how this applies to Mech building (The premade team thing would be for another post sometime in the future).

Think of your mechs overall weapons loadout as a team and you want to maximize how good you are. Now say you are not good at judging a hit or not hit with an AC20 before shooting. It packs a huge punch (an all star in the example above) but say either you haven't got the hang of it or your play style doesn't work well with it. Many would be stuck thinking "I need an AC20 because it is so amazing and it fits!" But is this really the best choice for your build or you as a player?

What I am trying to get people to think about here and also asking for help to do here (since I am not a math person myself) is push the stat analyzing we already have in part in the game. We have a mech stats and weapon stats tracker built into the game. It shows us everything from simple win/loss ratio and Kill/death ratio to percentage of hits for specific types of weapons. It doesn't cover everything (for example: my ER Large laser hit percentage says it is 65.7%, which I think is pretty good, but how many of those shots were full dmg? probably a lot fewer.) but what we have could certainly provide a start in analyzing what weapons you work best with.

So now I ask the math and statistically inclined among us for help. I want a program, app, spreadsheet or just a general rule of thumb that lets me use the stats that the game already tracks and provides to us in terms of telling me what would work best on a mech I am building. The goal would be that I have more than just a suspicion (even a well thought out and/or experienced one) or general consensus from experienced friends and forum posters that my mech will perform well before I go and play 10 or 50 or 100 games and decide maybe I need to tweak or even scrap the build.

Thanks in advance and I look forward to hearing people's thoughts on exploring this idea and solutions to it.

[scJazz]

To begin I'll respond with this... we do not actually have the tools. To do what you are looking for there needs to be at a minimum a Match Export Feature recording everything that exists in the client. THE DATA IS THERE! PGI for some unknown reason refuses to give us the button to cause the Client to record the data.

I have about 800 matches worth of screen shots... since the August 3rd ??? patch. However... stopping to write an application in any language that would extract the data in machine usable form is just too irritating. Again... the DATA IS THERE all that is required is PGI telling an engineer to spend ONE SINGLE DAY to record it in a usable format.

Direct reading of the Clients memory has been explored. But PGI has notified us that doing such is a TOS Violation and will result in Banhammer so we don't do it.

The best data we get in terms of performance comes from 12 v 12 Premade matches. They are the ones who end up hyper-optimizing their builds for maximum KDR. Still this is a data distortion since they all play with each other and are using voice. Setups that work for them are not necessarily transferable to pure PUG.

[Tiamat of the Sea]

To be perfectly honest, what the right weapons are on a given frame boils down more to what you want to do and are good at doing when you play than to the numbers on your weapons other than possibly your hit % with the SRMs, PPCs, Gauss, and Autocannon (hit % is much less significant with lasers and machine guns, which are both hitscan weapons).

No part of the game's current stat setup tracks whether you personally are better at predicting enemy pilot movement in confined spaces or open spaces, how well you can time shots while moving from cover to cover, whether you're good at selecting out of the way places to snipe from, if you have good instincts for infighting and spreading incoming damage, and so on and so forth.

Similarly, no part of the game's numbers other than range/damage/heat measurements (both over-time and per-shot for damage and heat) will tell you as an absolute what weapons go well together, and those aren't personal stats.

You're looking for an absolute solution to a wobbly 'problem'. It's like asking 'someone math for me whether zucchini goes better in rattatouille or zucchini boats'. Regardless of the math, some people like one or the other better or are better at cooking one or the other and no math is going to tell you about that.

As an in-game example, I seem to do really well when putting together medium lasers and an LB10-X, to the point that it supercedes my performance with an equal number of medium lasers and an AC/10. Is this demonstratable by my personal stats with the three weapons involved? No, not really, unless you look at my damage in matches where I'm using medium lasers and an LB-X and compare it to my damage in matches with medium lasers and an AC/10- and even that doesn't tell the whole story, because I've done a lot more of the other than the one recently, and a lot more of the one than the other overall, and have used LB-X autocannon without medium lasers a lot more than I've used AC/10s without medium lasers, and am still quite accurate with the AC/10 (indeed, just as much as with the LB 10-X by stats, despite the LB 10-X's shot spread). That said, I do terribly with medium lasers and an LB 10-X when I try to perform fire support rather than brawling, but can do quite nicely as fire support with an LB 10-X and a large laser and still brawl very well -too-.

The fact that I don't usually bring Large Lasers to a brawl means my hit % with them isn't necessarily indicative of how well I do at using them in such a situation.

In short: Wobbly variables and a not-entirely number based deciding factor mean this can't really be calculated. Relatively vague stat numbers (compared to what you would want to try to use) make it worse for trying to calculate.

Edited by Elli Gujar, 11 November 2013 - 09:54 AM.

[Burke IV]

I dont think you will find any clues to a good mech loadout in those stats. Its more about thinking out your role in the mech lab and then playing to that strength in game. I think it is to PGIs credit, that after messing about with a chassis often what "feels" right are "proper" loadouts (eg mix bag weapons, proper all over armour etc).

[Katotonic]

I agree completely with you guys. And I understand that there are innumerable difficulties involved here. Sadly, as it stands I realize that there is no definitive answer but what I am hoping for is a general pointer in the right direction based on numbers not feelings.

Elli Gujar your point on pilot skill not being accounted for in terms of what you do best could be partially covered by studiying K/D ratios on different maps. Or something like that anyways.

There are many things missing, for instance, I have played my catapult c1 as both a brawler SRM and pulse carrying mech and a LRMboat. But the stats as they stand right now do not specify which I was better in.

My point of the original post was that we should try to get as much of an indication as possible from the numbers, which while they won't be definitive might put us in a better position to be more successful down the line.

[Poptimus Rhyme Wallace]

IMO, there is enough data available to draw deductions and make an educated guess, sadly i know nothing of coding and do all my calculations on paper, so its not a service I offer, however since Katotonic is a member of the same Unit as me, I am currently in the process of evaluating his statistics as I post this...