i think you may be on to something. of course.


because we are dealing with an RNG as the main variable non of the predictions or estimations will have any practical meaning, but that doesn't mean it'd be any less cool.

 

actually, I was hoping that if we did it right, it would produce a positive or negative number in terms of the point surplus or point deficit the unit made over the course of the game.

Like I said, I'm not fluent in Mathese, so.............what's an RNG?

 

random number generator



too much in warhammer relies on a random number 1-6 or 2-12 to make any sort of conclusive statement.

best you can get is averages and probabilities.

 

Oh, yeah. Well... the way the numbers get imput, the point value assigned to, say an armor value is arbitrary. It is the best guess as to the amount of points an armor value will save on average. That way, the final solution will be an average.

 

that could be interesting. im not good at that kind of math. but good luck ill be interested to see what algorithms you come up with

 

Thanks for your input. I'll get the first draft of the equation posted and then I'd like to get it playtested and see if it holds up. Thats where the LO community will come in handy.

 

ok. Got the first(very) rough draft of the equation. Hope its not to hard to read.


E={[(S x 3)\(sv)][(S x I x 3)/(24)] + [(AW)(AW)-(OW)(OW)](10P)} - {[1.5s(0.75)][(BS)][(S)(S)-(T)(T)][10P]}

Where E is a positive or negative number representing how many points the unit made/lost,

s is the speed of the unit,

Sv is the save value,

I is the initiative,

AW is the units weapon skill,

OW is the opponents average weapon skill,

S is the units strength,

T is the units toughness,

P is the probability of an unsaved wound

BS is the units ballistic skill,

and n is the number of shots the unit fires (on average).

---------

I had to use my iPod touch to write this(its the middle of the night here) so I apologize for the weird things like writing (A)(A) instead of A squared- there's no superscript button on this thing.

Please let me know what you think.

 

very interseting.

its way tooo late for me to test this but ill give'r another look in the morning.

 

I don't believe this will be successful.

I'm doing 3rd year actuarial studies and the math, statistics, and probability I'm dealing with is far beyond what any normal person would like to know.

I think what you need to do is run some regressions to find the numbers you are looking for. I don't think we'll be able to crunch them out with some formulas. We'll have to run hundreds of games with many units and check each variable each game and plug the numbers in a regression, and do an analysis.

Let me explain. Say there are two guns, one is range 18" and the other is range 36". We know how well each gun kills marines, but we don't know how often it gets to shoot at marines, or any unit in particular.

What we can do is run hundreds of games, and basically count how many times each gun was in range and fired.

That way, we can put a weighting on each weapon depending on the chance it is in range.

I'd rather just use my subconscious memory from the over one hundred games I've played to answer whether a weapon's range is a problem. This is a better way to analyze units, math should only be used in answering questions that are specific and might be applicable often enough on the battlefield, making sure that the people that use the math know the limitations of the question.

 

Whilst I dont like maths hammer, Onlainari has done perhaps the seminal article in the finished tactica section on a smaller scale than what you are proposing.

I think I am sayign the same thing as him from a different angle, what you are doing is not viable because the modleing you are doing is not adiquate, there are too many varriables ( terrain, opponents army and unit selection, mission, special rules, vehicles, the composition of your own army) all of these you leave out of what is an already vastly simplified model, as such any number you churn out by turnuing the mangle handle on your equation is meaningless

the complexity of modleing this is so huge that to get even a remotely accurate value, you are better doing it by observation than statistical forecast

 

Is it possible to create an accurate mathematical model for an entire game of 40k? Yes. Clearly it must be.

On the other hand, the volume of input and the complexity of the model would be vast. Your suggested formula only just barely begins to capture the sort of thing it would need to include. It wouldn't quite be trying to predict the weather, but I still can't imagine that such a project could be done without far, far more work than it would be worth--and, basically, it would take someone with a significant amount of expertise in the area.

It's not an impossible idea, but it's almost certainly a bad idea.

Now, what would be far more interesting would be to write a genetic program which would sit around playing games of 40k against itself, tracking the results, and using those results to fine-tune its own lists and play decisions. This would also be difficult and require a not-insignificant dergree of expertise, but the results would almost certainly be more interesting.

 

I work for the Pesticides Safety Directorate here in York and part of our work is analysing data for Pesticides (safe usage regimes, maximum residue levels, efficacy etc etc etc). You would not believe how much data is generated in the simplest of tests.
For a simple pesticide approval the data package can run to 5000 sheets of data and often can run to many many times more than that.
I can tell you right here and know that it would be nigh on impossible for a laymen to create a working mathematical model of a single game of 40K. The sheer amount of variables is staggering The military run theoretical war games and they use a staggering amount of computer resources and highly qualified people in the field of statistics to get run even the simplest of simulations.

As for the genetic program, we have only just got to a point where computers can consistently beat the best Chess champions and I think a game of 40K would only be slightly less challenging to write a program for. Just the sheer amount of variables thrown up by the simple expedient of the sheer number of army list and unit choices available would make it a nightmare. It would be cool but very hard to do.

 

I can tell you right here and know that it would be nigh on impossible for a laymen to create a working mathematical model of a single game of 40K. The sheer amount of variables is staggering The military run theoretical war games and they use a staggering amount of computer resources and highly qualified people in the field of statistics to get run even the simplest of simulations.

Absolutely. Like I said, it's possible in a sort of theoretical manner, but its implementation would be so far beyond anything I can even really imagine that I can't view it as a wise undertaking. If someone had the expertise to pull off a task like this, they'd certainly have more profitable and fulfilling things to be doing with their time than making a model of 40k.

About the neural net, though, it has been done. I've been looking for the source, and I can't find it, but I read an article about a year ago about a fellow who built himself a neural net to play in Traveller tournaments. It won. The fleets and game-plans it devised so utterly dominated the tournament scene that the fellow was eventually banned from playing in them. Good naturedly, of course.

Anyway, that is also an expertise-intensive route to take. I'm betting, though, that the results of a well-constructed neural net would be both somewhat easier to obtain and somewhat more interesting when obtained.

It's not a very important point, though, since I don't know anyone who could create either program.

Anyway, the point was mostly to tell the OP that, while his goal isn't actually unattainable, it basically won't resemble, at all, what he has proposed so far.

 

Well, I'm going to throw in my ten pence here:
It, as stated by others is possible, but not worth it
My reasoning is that the "generalship" - not being used in the equation is such a large part of the equation - especially in troop choice.
WARNING - LOOK AWAY IF YOU HAVE MORALS
Maybe it's as i'm a ex-CCG card player (4th in the European finals, the year after 17th, while coming in the top 8 in the uk 2 years out of 4, blah blah blah i could bore you) but i wholly beleive 3/4 of a competetive battle is in the army list.
- now I'm going to wash my hands for typing "some thing to do with the m-word"
- by the way kids, don't do it. It's not big, and it's not clever.
I would like to apologise and thank everyone who ever listens to me for their patience - there is a inner demon inside me now that can't help but try to break rules (in order to abuse them) and i sometimes need to be reminded to let it go and enjoy the game. (It's Probrably for the best i've lost the last 7 out of ten games - in my book that shows signs of rehabilitation.)
OK- ITS SAFE NOW

What could be useful is writing up some crib sheets for effectiveness for various units, against the standard units.

IE Marines vs marines

You'd do a calculation of to hit, (once at 35", once at 23", once at 11") composed with: to wound, then saves (vs standard marine) for points over kills (three results)

A calculation of Melee (once when charging, then once when "static", and a calculation for equal init (cover)) with amount of attacks, then to wound, then save in initiative order (or not as the case may be) resulting in three results (maybe with notes i.e. instant kill etc)

and (with exel) pump out various units by just putting them into the exel spreadsheet.
It wouldn't be that hard to slot in various weapons, to get your results, and you could do a vehicle chart (I did mathammer once for all the IG weaponry vs Armour values when bored on a 3hr train journey.)

etc etc

However, for a vanilla standard marine, you're going to have to do ? 10 basic troops for the different armys? 3 different types per army? (for marines i'd look at assault, tac and termies), and then that's a hell of alot of work for something you probably know already if you just look at it - is it a good or bad matchup?

Your experience will tell you all you need to know, you're far better playing more games - I don't think it'd be a effective use of your time. No offence - but it won't be that useable. What happens with an opponent that infiltrates? Deep strikes? fleets? has no armour? has all armour? deploys after you? will you do a different list for if they go first?
Not trying to be negative, great idea but my god man don't do it - work out the ending of pi and get rich.

 

Attempting to create a formula to come up with a number representing a specific unit's overall effectiveness is not only inane but its a moot point.

Its a moot point because GW has already included a ballpark rating of a unit's effectiveness in the points system. Marines cost roughly 3 times what a IG soldier does because a marine is roughly 3 times more effective. Terminators cost roughly 3 times what a standard marine costs because they are roughly 3 times more effective. Terminators are roughly 6 times more effective than an IG. The pts cost are rough estimates because their pts values weren't the result of an equation where the units stats were plugged in, but rather the result of thousands of test games over a couple decades by people who spend at least 8 hours a day thinking about very little other than game mechanics. That produces a pretty thorough estimation of what theyre worth. We all know which units are worth more or less than their points values because those are the ones we complain about constantly.

Its inane because different units are good against different targets and weak against others, and different players know how to use them more or less efficiently. No matter how advanced your model is, mathematics has still never been able to properly address human skill or error into the equation.

 

Playing the game is the only reliable way to determine what works, mathhammer is only of very very limited use and can give you a limited overview of how a unit 'may' work within a limited range of variables or in a very defined siuation.

The only way to truly become aware of what works is to play the game. The sooner that this simple concept is realised the better, we have had many attempts to rationalise or to create a single working model/system (mathhammer) so anyone can say "oh this unit/list/army etc will definitely work" and not one of them has ever worked as well as simply putting the models on the table and rolling the die.

Mathhammer can give you a very simplified limited idea of what may be possible with a weapon or units firepower etc but that is it. If you want to know what works 'play the game'

 

Rikimaru is right to a degree. While mathmatical analysis of a unit's abilities and attributes can yield a lot of very specific information (how likely is this unit to kill this other unit in this particular scenario?) producing a complete overview and evaluation of a unit would require analyzing hundreds of thousands of different scenarios and creating a system of weights so that the results of each scenario impact the total result no more than the likelihood and importance of the scenario warrant--and calculating values such as the likelihood or importance of a particular scenario is not something which can be done with probability math and unit attributes alone.

That, really, is the problem--and the place where playtesting becomes important. Playtesting will never give you a better idea of how a unit will perform in a given scenario than will "mathhamer." It just can't. Mathhammer amounts to playing out a particular scenario an infinite number of times--using a virtual set of data which is far more significant than any set of data you could possibly gain through experience.

What playtesting is good for is gaining the data necessary to weight the importance of particular scenarios--to inform your decisions regarding which scenarios to math-hammer out.

Playtesting is not the be-all end-all of analysis in a table-top game. In that assertion, Rikimaru is wrong. It will routinely fail to provide data on specific scenarios which is as accurate as the data which can be obtained through mathmatical analysis. Mathmatical analysis, though, is useless if you can't pick the right situations to analyze--and picking those situations is not something you're capable of doing through math-hammer. It requires experience.

A model such as the OP is discussing is possible. It could be created. It would take far more work, though, than slapping together a haphazard handful of formulas. It would require, in addition, countless hours of play to gather the data needed to properly weight the results of the multitudinous (and we're talking, once again, in the hundreds of thousands) of bits of mathmatical analysis.

 

I have never yet once met a player who said " I won that game because I used math hammer to predict every single variation and variable of play that could arise and had the counter to each variation and variable prepared in advance' bow before my math hammer greatness you pitiful dribbling non math hammer using wretch". I have also yet to meet any player that has based an army list purely on math hammer. I do however continually meet players who have refined lists into veritable killing machines through playing games and refining the list using play testing.

Now I am not saying math hammer has its (limited) uses; however knowing how a unit should perform with its shooting or how vulnerable it should be to failing saves (probably the two most common uses of math hammer) more often than not lacks any similarity to how the unit performs within the complex system that makes up the 40K game.

The only way to get an accurate picture of how a unit will perform within the army list and against the myriad of opponent types we face in 40K is quite simply to play the game.

As far as your 'specific scenario' point goes, well I am only interested in the scenario's I play and I get all the data I need from playing said scenarios. I can apply the experience gained from similar scenarios or simply by knowing what my list is capable of from experience of playing it to any new scenario.

Lets face it Left you have basically eluded to the fact that trying to figure out every single scenario you MAY face is next to impossible and you then have to figure unquantifiable things in to the mix like your opponent plays crap because he has a hangover for example or you play a rushed last two turns because the tournie time limit is nearing.

Play testing is the be all and end all dude, because at the end of the day that is what it all comes down to. You can sit at your computer for years and work out how theoretically the game should play out, but at the end of the day you have to sit down and play the game and then your suddenly thinking "why is this game not going as my formula predicted". Nothing comes close to educating the player about what is effective more then actually playing the game.

We all use math hammer to point out that a unit should kill 2.6 Marines a turn (or similar) and I have done myself and there is nothing really wrong with that. However solely using that limited info to build winning lists or creating a math hammer system to create the perfect list is just never going to happen..

I tell you what, you create a list purely using math hammer, using units and configurations that math hammer says should be optimal (do not let your opinion sway the list, just use math hammer) and I will produce a list using my experience (no math hammer), we could post them, let some others use them and then see which does better. I am confident I know which will do better, are you?

Oh and one last point, it is impossible to carry out an "infinite" amount of tests (if its infinite you never get a result do you?) and unless you account for every single conceivable variation or variable then the results will be inherently inaccurate. At least with play testing the result you get uses all the variables applicable to that situation, I/E it is impossible to miss out a variable because the situation is complete unto itself.

 

Haven't played me yet have you?

Sorry couldn't help that one. But let me guarantee you here and now, Mr Maru, that if we ever play and I win I will repeat the words below Inigo-Style

"I won that game because I used math hammer to predict every single variation and variable of play that could arise and had the counter to each variation and variable prepared in advance' bow before my math hammer greatness you pitiful dribbling non math hammer using wretch... My name is Math Hammer, you killed my father, prepare to die!"

Ahem, I may have added a bit to reinforce my references.

Oh, and as for the thread, the idea is ridiculous. Give it up.

 

A Mercy Killing ...

Oh, and as for the thread, the idea is ridiculous. Give it up.

D'accord. It was dead and buried and then someone went and dug up the corpse. We can't have mouldering undead threads stumbling around these Boards (there's quite enough of that over on the FB Boards) so I think now is an apposite time to put this one away for good.

Go gentle into that quiet night ... ...

Thread Closed.

~ Raven ~