I was wrong all this time with the Stalker

[Doctor Machete]

Well, after all the time I've been spending talking about the Stalker being the absolute best brawler, the mvp, I recently discovered I was wrong, not totally but still big time.

Recently I decided to revamp my setup of planner plus chances calculator to remove some code I don't use anymore and add a few QoL extras (like auto assign weapons/equipment to slots, so I can check in a fast way if everything fits), and after I finished I decided it would be nice to add an alternate method to calculate chances using random number generation (instead of calculating the final chances) with the hope that perhaps would be faster. When I did it, I discovered that my SRM (and MG) code calculation was wrong, it didn't factor correctly diminishing returns. I investigated how could that happen, as I was pretty sure it was working right (and I still have the test sheets used for to compare results), because I did a lot when I first made it. I managed to track the error to one of the many performance upgrades that I did long time ago, when that was a big issue for me, as a a 5-6 mech calculation could take easily near two minutes (much more if two of them were LRM boats) and that being the reason I also had an specific mode for simplifying LRMs, which could easily exceed the maximum time allowed. Also ironically the mistake was due to overconfidence, because it is quite dumb, just two words in a bunch of code, while the LRM part, which took me a lot more time to get it into decent performance and I've redone it from scratch several times trying to get better performance, is working fine. Perhaps because that being the case I tested it more thoroughly each time I made changes, and not so much when I dealt with the easier parts. Becase the SRM code is fairly easy but it's not simple or fast to test that you can check with a binomial calculator like I can for single hit weapons.


So yes, I've been a fool, while I've had a mountain of data from logs I never bother to look into this specific issue after the first testing, as I was sure it was right, showing how easy is to keep believing what one already believes, but it was an honest mistake and while embarrassed of how long took me to discover it at least it didn't need anybody else to point it out.

Now, the correct numbers would be (for the usual comparison):

            SINGLE SALVO                    DUAL SALVO
 DR     2xAC20++  6xML++/4xSRM6++     2xAC20++  6xML++/4xSRM6++
 0%       32%         39%               54%         83%   
20%       32%         29%               54%         76%
40%       32%         13%               54%         59%
60%        3%          1%               15%         24%

So still quite decent single salvo and very good dual salvo but not so much a powerhouse as I thought. That dual performance no doubt helped me believing it, because I almost never PS in an isolated way but almost always in couples, be it from the same mech in a row, two same mechs or a single Stalker plus a long range mech (LRM or direct damage).

Probably I'll add a comparison screenshot of several more builds, so I'll bump the thread enough time for regulars with who I've got many discussions will notice my mistake and they know I'm wrong in this.

Note that while changes my opinion in some aspects it doesn't in others, like the multi+breaching recurrent theme of discussion.

Here some more setups:

Spoiler

 

[Stuckenschmidt]

That may be true (and probably is), but the Stalker, just like all chicken-leg Mechs, will always have a special place in my heart and Mechbay.

 

[Shark7]

I find the Stalker to be one of the best missile boats in the game. SRM or LRM. And even the stock configuration is just scary.

 

[Doctor Machete]

Updated with more builds.

 

[Icewraith]

Curious- what if you switched the 6LL/1AC5 to 4LL/2AC/5?

Also 2LL/3ML/6SL GRH.

4LL/4SRM2 or 4LL/4SRM4 Stalker.

 

[Doctor Machete]

For the A-2 with extra AC5 logically gets better but hugely better and at the expense of a lot of armor, and don't forget you can't use Gyro. The LL versions of the Stalker get lower chances, and lower or much lower armor. If you really want LLs then imo better just add some kind of combination of AC2/AC5, with even less chances but you can do all the damage at LL range and get full benefit from a Rangefinder++, that's why I add the KC-LLAC again, with 2xAC2s but 1xAC5 also is fine, perhaps even 1xAC5 1xAC2, 2xAC5 is too much.

The Grasshopper with LLs is more of the same. You get slightly lower chances but sacrificing a lot of armor/heat and you'll need all the armor you can get in order to fire at SL range.

Spoiler

 

[Icewraith]

For the A-2 with extra AC5 logically gets better but hugely better and at the expense of a lot of armor, and don't forget you can't use Gyro. The LL versions of the Stalker get lower chances, and lower or much lower armor. If you really want LLs then imo better just add some kind of combination of AC2/AC5, with even less chances but you can do all the damage at LL range and get full benefit from a Rangefinder++, that's why I add the KC-LLAC again, with 2xAC2s but 1xAC5 also is fine, perhaps even 1xAC5 1xAC2, 2xAC5 is too much.

The Grasshopper with LLs is more of the same. You get slightly lower chances but sacrificing a lot of armor/heat and you'll need all the armor you can get in order to fire at SL range.

Spoiler

I’m a bit confused how you built the A2 4LL/2AC5. In theory swapping 2LL for 1 AC5 should be ~tonnage neutral only you have a much lower heat burden.

 

[Doctor Machete]

I’m a bit confused how you built the A2 4LL/2AC5. In theory swapping 2LL for 1 AC5 should be ~tonnage neutral only you have a much lower heat burden.

Oh, my mistake. I understood 6xLL 2xAC5, because I wouldn't consider an Atlas-II with just 4xLL, even if I have to sacrifice the Gyro. Now it makes sense XDDD. Later I'll upload a roundup of several LL/AC5/AC2/PPC setups with the A-II. With the A-II what I always tend to do is using setups not possible with other mechs, even if they're not optimal. Like the 5xPPC, which is my favorite setup of the game, although a 6xLL +something is better imo.

 

[Icewraith]

Oh, my mistake. I understood 6xLL 2xAC5, because I wouldn't consider an Atlas-II with just 4xLL, even if I have to sacrifice the Gyro. Now it makes sense XDDD. Later I'll upload a roundup of several LL/AC5/AC2/PPC setups with the A-II. With the A-II what I always tend to do is using setups not possible with other mechs, even if they're not optimal. Like the 5xPPC, which is my favorite setup of the game, although a 6xLL +something is better imo.

The AC/5++ has the magical 55 damage to take over half the head armor at 40%DR. If you don’t land the AC/5 on the 6LL/1AC build you need at least 3 LL to hit. Headshot-wise I’m pretty sure 4LL/2AC has roughly the same potential at 40% with a much lower heat burden, and it can use a gyro.

If you’re taking requests, I’m curious how the GRH and STK builds fare without any++ weapons and with 2 of each ++ weapon and the rest regular.

 

[Doctor Machete]

The AC/5++ has the magical 55 damage to take over half the head armor at 40%DR. If you don’t land the AC/5 on the 6LL/1AC build you need at least 3 LL to hit. Headshot-wise I’m pretty sure 4LL/2AC has roughly the same potential at 40% with a much lower heat burden, and it can use a gyro.

Yep, that's what happen. But being a long range mech you can afford to sacrifice some armor for a bit more punch, if it's not too much. And also the performance for CT core and vehicle killing will suffer a lot, although vehicles are not shown (more total damage becomes better), but you should take into account they have less hit locations and most of the time you probably don't want to spend PS on them unless there are no mechs around.

Spoiler

If you’re taking requests, I’m curious how the GRH and STK builds fare without any++ weapons and with 2 of each ++ weapon and the rest regular.

Spoiler

And btw, if you're familiarized with spreadsheets and simple formulas you can very easily check the chances for setups with small number of hits, it just take some time, because 6-7 hits is not super fast to do it but still manageable in a table by hand (copy/paste), resulting in 64-128 rows (remember most of it is copy paste of values and formulas). In this case you can set very specific conditions, like if A or B weapon hits and at least two hits between A+B+C+D+E+F.

Spoiler

Edit: graph colors were messed up because I copy pasted whole row values instead of selecting each setup one by one. I'll generate and upload again.

Edit.2: colors are right (actually already were right the previous time I updated)

 

[Icewraith]

Granted you’re addressing a far more general case than “will this combination of weapons reliably headshot”, but it seems like you could be doing a lot more with several sets of binomial functions for a lot less space with your check sheet.

 

[Doctor Machete]

Granted you’re addressing a far more general case than “will this combination of weapons reliably headshot”, but it seems like you could be doing a lot more with several sets of binomial functions for a lot less space with your check sheet.

Sure, in that example you could isolate from LLs into to one binomial and use another for AC5s, but as posted it is very straightforward and easy to visualize, provided the dataset is small enough. And you don't need to manually copy paste, you can make a function to generate all of it, including the sheet itself, formatting and formulas.