Page 1 of 1

How good is the AI at signing its own RFAs?

Posted: Thu Feb 16, 2017 8:18 pm
by nino33
Hello everyone :-)


As the thread title says, I'm wondering....how is the AI at signing it's own RFAs? Are Good-Top players being let go for nothing?

This has been brought up before, it seems some improvement has been made over the last year, and I'm wondering how people are finding it now...

Today on the HFBoards EHM Thread it was discussed (see posts 911-916) http://hfboards.hockeysfuture.com/showt ... t128233271 and a link was provided to a "RFA's non-tendered ---> UFA status" thread from March 2016 on the SI EHM Bugs Forum (but there were no saved game examples uploaded, nor did Riz provides any reply/explanation)

So, please chime in on the issue if you can!
If you think there's still; an issue I think it'd be great if you could post in the SI Thread and (importantly) upload a saved game example for Riz to look at (I don't know, but it might be helpful to have one in earlier/later June and July 1st) - I know it can be a pain to post/upload saves, but such efforts can really improve the game!


Thanks for your time everyone :-)

Re: How good is the AI at signing its own RFAs?

Posted: Fri Feb 17, 2017 10:35 am
by Shindigs
On a scale of 1-10, I'd give the AI a solid -2 on their ability to sign RFAs. The general rule of thumb is that they will play their prospects in the AHL way too long, their rolled contract ends. And since the AI thinks the player is still an AHL player, they refuse to give them that very reasonable 2.5M bridge contract. The player then goes to the SHL, Liiga, DEL or KHL. Generally they will stay there until they are around 27 and their rep has gotten high enough that the AI considers them an NHL level player. At that point the players will rejoin the NHL at their peak. This is a consistent thing is all saves, so in this case supplying a specific save game seems a somewhat moot point since Riz could make one himself by just starting a fresh save and sim 1 season forward over night. Since it always happens. Especially with goalies. John Gibson in the DEL is becoming a mainstay of all my saves at this point. Guess he doesn't like Russia for some reason:S From my point of view the issue with RFA management doesn't stem from the trade AI itself so much as the AI putting way too much weight in a players reputation value (in all aspects of the game) rather than their actual ability. It would be interesting to see how a db with all prospects rep manually set to be higher would play out. But I'm not inclined to spend the time to make that.

Actually now that I think about, making the AHL/CHL grant players more rep would probably be the real fix. Either that or dropping the coefficient on rep in AI scripting. But that might make older players phase out of the NHL too soon. It's hard to say when you can't see how it's actually coded, and only going off how it works in practice.

Re: How good is the AI at signing its own RFAs?

Posted: Fri Feb 17, 2017 4:25 pm
by nino33
Here's an HFBoards conversation from today that Riz responded too, and he also noted the RFA issue so I'm posting the conversation here (the red font I added)
Here's a link to the actual thread http://hfboards.hockeysfuture.com/showt ... 99&page=37

Smirnov2Chistov wrote:The replay factor is a big deal, plus if you challenge yourself to try and rebuild the team it's a different way of looking at the game. (Signing Chinese players, Spanish players, etc.)

Yes. You can exploit the game like its Football Manager (sign all the up and coming potential superstars), but it just doesn't even come to mind like that to certain players. (Nothing against you)
Lebowski wrote:But as I said... I'm not exploiting the game. I barely even touch the trade and FA features. I'll sometimes trade NHL waiver eligible players that can't crack my roster for whatever picks I can get, and that's if I can get any. Otherwise, it's all about drafting right and getting that consistent influx of talent. The AI lose so many players as free agents every year that they're always left behind since they can't fill up the holes left by superstars leaving every year... Which is only natural.

If you can resign your top players to moderate salaries and draft somewhat decently, in my experience, you usually end up being unbeatable within 5 years. And again, that's without the exploits.

I know about the TBL challenges, I know these sort of things are non issues against human players... But I think what Riz should strive for is to make a game that forces the player to use any ressource at his disposal to remain competitive, because that's the way it is in real life.

Sure, you can always force yourself to play a certain way under made-up rules to make it more challenging... But to me it only goes to show how broken the game is in the first place.
Lebowski wrote:But I think what Riz should strive for is to make a game that forces the player to use any ressource at his disposal to remain competitive, because that's the way it is in real life.
Riz wrote:Yes, that is the aim believe me. However coding the AI managers to be as intelligent as your average human armchair GMs takes a lot of time and finetuning.

With every new feature added to the game, there is typically an AI element that needs to be added to make sure the AI can do similar evaluations the human does. And finetuning these can be tricky at times, especially when it comes to player evaluations for offering new contracts and the status of the player within the team depth chart. Typically the human managers look at the actual player production a LOT more than the AI managers, who rely more on analysis of the actual player ability. And when players sometime outperform their abilities and have a good year statistically (for example playing on a line with better players), it looks weird when the AI does not appreciate that production by the pure numbers and does not re-sign someone who had a good season. This is an area that I'd like to improve in the future but like I said, these things take time to improve as a lot of analysis and testing is needed.

As was noted above, if you find such cases with RFA's for example in the game, the best bet to get my attention is to post on the SI bugs forum and upload a saved game from around a week before the player goes RFA and the team hasn't re-signed him. I'm working on FM mainly but have some time for EHM on the side, so it may take some time occasionally for me to reply to posts on the bugs forum but a post + save makes it much more likely for an issue to be analysed, identified and resolved.

Currently most of my EHM time is going towards the built-in editor work, so I'm not checking the bugs forum as frequently as when working on the gameplay issues. But I'll do another purge there at some point again when I've reached the next milestone in the editor work. The 1.5 update will include gameplay improvements as well, just like the previous updates, so any issues logged on the official bugs forum with saves uploaded on our FTP will help improve the gameplay further.

Re: How good is the AI at signing its own RFAs?

Posted: Fri Feb 17, 2017 6:47 pm
by Shindigs
Riz wrote:Typically the human managers look at the actual player production a LOT more than the AI managers, who rely more on analysis of the actual player ability.
Huh, I'm really surprised that he brought up the AI not valuing production that much. Had this been before my Jets save I'd have nodded in agreement. But this makes why in gods name Pittsburgh decided to give me a 1st rounder and 2nd rounder for Burmistrov who was on a 7 year 4.8M a year deal all the more perplexing. I guess maybe his rep got inflated by the cups he won?