The "usefulness" of the Deflection Attribute

[Shindigs]

I've been thinking about how useless this attribute is lately, partly because the 2D engine doesn't have any animation for it. So you'd never see one even if it happened. So I just went through the events_eng.cfg file to check what all the events in the Play-by-Play related to deflections are. There are a lot of them, several that make limited amounts of sense. But that is beside the point.

I then went through a full month of games (the ratios of the stats remained nearly static throughout the month, so I didn't see much need to do the whole season, since that would take forever)

As expected the conclusion was that you can ignore the deflection attribute, as well as the "scores from rebounds and deflections" scout snippet. Because a Whopping 1.69% of the goals scored were from players deflecting the puck in. Actually the best attribute for scoring from deflections isn't to be found on your players at all. It's on the oppositions defensemen, since they scored 7 times as many own goals from failed blocking attempts as the forwards did from the super impressive 21 attempted deflections across 760 shots. To sum up the findings from this limited sample size test:

Shots Attempted: 760
Shots deflected by defensemen: 78
Own goals from shots deflected by defensemen: 7

Shots deflected by forwards: 21
Goals from shots deflected by forwards: 1

If this were to remain across a much larger sample size, which it probably would do within a reasonable margin for error, you'd see something like 4 goals from deflections every 250 goals (Caps scored 248 this season) I couldn't find any stats readily avaliable for how many of those were deflections, but I think it's fair to say it was more than 4; since you won't be chocked to see a single NHL game with more deflection goals than 4.

Some funny things highlighting how well the match engine works were:

In one game, the same shot was deflected by the same Dman 2 times, a few games later the same shot was deflected by one Dman 3(!) times. That's some solid physics right there.
In another game a player fired a backhander, only to then deflect his own backhander in the following second. Meaning that he presumably shot it so slow he outskated it and then deflected it...or something. Working as intended!
Finally the singular goal scored from a deflection wasn't given to the player who deflected it. It was given to the original shooter, meaning I can't even prove with 100% certainty that a single deflection was scored. But the final event in the Play-by-Play before the goal was "The shot looked to have taken a deflection off <pl>" which is tied to the EVENT_RICOCHET_SHOT in events_eng.cfg so it probably means that even if a deflection goal is scored, it won't officially count as one anyways.

So in conclusion this means that I was right in feeling a bit hard done by whenever I saw a player gain +1 from Shooting training only to realize it was +1 Deflection. Almost as fun as when someone who isn't a Center gains +1 Faceoff from their Off. Skill training.

Naturally the samplesize is far, far too small to be irrefutable proof that the attribute is absolute trash. But it would need to improve so much over the remaining months to actually become useful. At just over 2 attempted deflections per game, even if a 20 deflection player would have a 100% chance to score from them (which he clearly won't) the chance of him being the person attempting them isn't exactly something you'd bet on. Very roughly estimated on a forward with 15min ice time he'd be on the ice for 1/4 of those deflections, and there'd be a 1/3 (3 forwards) that he be the one in position to actually deflect it, you end up with in the region of a 1/12 chance that the deflection attribute will even be used AT ALL. Also note that those 2 attempted deflections per game are for both teams combined. So he'd have a 1/12 chance to get to use the deflection attribute once per game. At that point even with 100% chance to score from it, the attribute would just be too insignificant to matter. Across an 82 game NHL season you'd look at about 7 attempts to deflect for that player. And since 1 deflection isn't a 0% chance to succeed, and 20 isn't a 100%, whatever difference in deflection there would be in between the two would ultimately be pretty pointless. Sure all goals help, but wouldn't you rather have that attribute point in Wristshot or Slapshot, you know those attributes your forwards actually use multiple times per game.

If anyone has had any luck using some form of tactical setup with funnel on the PP and overload slot with some kind of players that really excel at deflections, I'd love to hear about it. But really, it looks like this part of the game engine needs some love.

[CJ]

Good researching!

[nino33]

Are you playing Quick Sim or Full Sim?
Are you controlling a team at all?
Have you looked at some NHL games for comparison?

I'd be interested in seeing if you get the same results with power forwards personal tactics set to always crash the net and team tactics set up to maximize deflection possibilities (I'd think crash the net and/or point shots)

I recall seeing deflections in the 2D myself, but that was with EHM07...
And I recall reading about having a C with high deflections being effective on the PP, but that wasn't with EHM1 either so maybe it's different now

[marksbros6]

Don't usually screen the goalie on the PP in my EPL games but remember one time on EHM1 that I forgot to change the PP instructions and Markko Tahtinen scored a good few deflections in the 2D view with the default PP tactic from playing Centre.

[Shindigs]

Yeah, I stopped using 2D due to how mad its "shortcomings" make me. I've seen a fair few things that "look" like deflections, but since the game credits the goal to the shooter, not the deflector, its so hard to tell. The play-by-play is more than a little cryptic at times too, but it does give more reliable information than the 2D at least.

Also the games showed was in the EIHL, using almost exclusively finesse forwards, having the direct attacking AI coach deal with all lineups and settings. They were all my team's games, so it should be full sim. Almost all the deflection own goals were from the opposing team, but the only probable deflection goal was also from the opposing team.

I play all games with hidden attributes, no searching out players by attributes, limited scouting ranges per scout, AI coach handling all roster/tactics. Limiting the amount of players I drop to free agency cause they didn't set the world alight. Trying for a somewhat realistic approach, but it does make the testing way less scientific due to all the uncontrolled outside influences. But then there are so many unknowns in any EHM testing anyways, due to not knowing the exact nuts and bolts of the match engine.

The thing about this is that you couldn't actually fix this issue without getting back the "too many goals in the NHL" issue that seemed to be the vast majority of the EA feedback from the community at large. I have no doubt there was some much more useful stuff coming from here, but I'm not inclined to go read old threads to find out:)

To actually up the amount of deflection goals to the somewhat ridiculous levels in the actual NHL these days, you'd also need to tone down other ways of scoring in EHM, since the amount of goals is about right. It's just what type being scored that is wrong. It's pretty safe to say that the cross-crease one-timer goal is more than a little overrepresented. Mostly due to the players actually being dots, unless the pass goes smack through the middle of the dot, the pass wont be intercepted. Real players aren't dots, they have legs and sticks to take out passing lanes quite efficiently. If this was better represented in the match engine you could start looking at toning up deflection goals. But until that is done, actually "fixing" this, assuming it does need fixing, would make it so the scorelines got all kinds of funny again.

edit:Oh and another funny thing, that more or less sells me on the idea that EHM is a modified version of an old FM. If you look through events_eng.cfg on line 1369 you find this line EVENT_TAKES_BALL_TO_LEFT, seems riz forgot to change ball to puck in that line
edit2: and a few lines later you have EVENT_TAKES_BALL_TO_RIGHT...well then.
To clarify the EVENT_GOOD_INTERCEPT does exist, and it returns a string of text for when a pass, rather than a shot, is intercepted. The issue that in 9 games I never saw it once, which means its not going off like it should. I don't care how good you are at passing, you won't have your whole team go 9 games without intercepts, especially not when the majority of the team has a 7 or worse in passing. The reason pass completion is as low as it is seems to stem more from actual missed passes, that often go to icing calls, more so than intercepts.

Scrolling through the cfg there are actually an absolutely enormous amount of different strings for different events, but I can only attest to seeing maybe like 5 of them repeated over and over across those 9 games. Either every single team in the EIHL plays the exact same, very 1 dimensional hockey. Or the Play-by-play is broken, or even worse the match engine itself isnt having a lot of these events fire due to some programming whoopsies.

[nino33]

, using almost exclusively finesse forwards, having the direct attacking AI coach deal with all lineups and settings

Not a setup that's likely to get many deflections...

They were all my team's games, so it should be full sim.

In game, click on the gear at the top of the screen, then Detail Level...this is where you select Full Sim (Full Detail) or Quick Sim

But then there are so many unknowns in any EHM testing anyways, due to not knowing the exact nuts and bolts of the match engine.

IMO there's lots you can still learn without getting to the nuts and bolts of the match engine (and that knowledge would help in the understanding of the match engine too)

The thing about this is that you couldn't actually fix this issue without getting back the "too many goals in the NHL" issue that seemed to be the vast majority of the EA feedback from the community at large. I have no doubt there was some much more useful stuff coming from here, but I'm not inclined to go read old threads to find out:)To actually up the amount of deflection goals to the somewhat ridiculous levels in the actual NHL these days, you'd also need to tone down other ways of scoring in EHM, since the amount of goals is about right.

The scoring can't be both "to high" and "about right"
FYI the to many goals issue was with EHM07 (and fixed years ago by a user created patch); I've not heard of any concerns about high scoring with EHM:EA/EHM1

edit:Oh and another funny thing, that more or less sells me on the idea that EHM is a modified version of an old FM. If you look through events_eng.cfg on line 1369 you find this line

I wouldn't think some shared code would automatically for sure mean a modified version of an old FM

Either every single team in the EIHL plays the exact same, very 1 dimensional hockey. Or the Play-by-play is broken, or even worse the match engine itself isnt having a lot of these events fire due to some programming whoopsies.

For me there's nowhere near enough understanding/evidence for making such statements...


I do think what you're looking into seems immensely cool/interesting! I've never seen anyone ever reference the events_eng.cfg details before!
But that means to me I'd want to be cautious about making definitive statements at this time

[Shindigs]

Yeah, most of this is just me thinking aloud. But the complete lack of some of these events should, logically, mean that something isn't working right. Hopefully it's just an issue with the Play-by-Play. Since it seems somewhat rare people even look at it, it would be easy for some bugs to sneak through there. But if the issue is with the match engine itself not firing all events it could/should that is a pretty big issue.

The "too many goals in the NHL" is mostly from the steam community feedback, and I have seen it for EHM:EA on here as well, but I can't remember where since it was months ago.

Also on the note of the EHM engine being a modded old FM, I remembered where I got that from. It was Jack "WorkTheSpace" Peachman's YT playthrough of EHM:EA, if memory serves me right one of his first reactions when loading up the game was "oh, this looks like it's built off an old FM engine". And since he used to work in QA on FM at SI, I feel like he should know it when he sees it.

Presumably the same was true for the older EHMs as well, when you're working solo on a project of this size, all possible shortcuts need to be taken, and building your own game engine from scratch isn't exactly a shortcut. Especially not when your company already have an engine that is made for a quite similar game, since most of EHM and FM are the same. You're doing the same things, just with different coats of pain on it.

On the scoring issue, my wording was "getting back to ..." as in, it used to be an issue but isn't any more. But you can't tone up the amount of deflections goals, with the current amount of other goal types, without getting back to a state where you have too many goals being scored. At the moment you have a reasonable amount of goals being scored, maybe a bit high in some of the euro leagues at least, but it's not too silly. If deflection goals were upped to a more realistic level, it would put the goal total way too high. So something else needs to be toned down first, or upping the deflections to a more realistic level would serve to break the game more, which isn't the end goal for anyone (hopefully).

And I do agree that my specific setup won't lead to a lot of deflections. But I am counting my opponents too, and most teams in the EIHL are mostly made up of ex-AHL and ex-ECHL checking/power/flashy/Quick/Skilled forwards, which are the more physical type players that you would think should get a lot of deflections in. That only 1 in 21 deflections were actually scored means that even if you do get in a deflection, which they don't nearly enough, it's way too easy for the goalies to save them.

If someone else is playing a more physical style hockey than I am, and either know a way to automate looking for these events in the Play-by-Play or have enough spare time/sanity to count them in another league with a bigger sample size, that would go a long way towards getting a more complete image of how the deflections do/don't work. I just do my testing as I play normally to sate my curiosity. I don't much feel like trying to perfect a tactic for a playstyle I don't play, in a league I don't play just to find out if an attribute I don't rate is good for a style I don't play. I'm just not altruistic enough for that I guess.

Edit: After the latest patches there are a lot more deflections. Still the vast majority of them aren't goals (and most goals are still own-goals, but not counted as such), but at least the deflections do happen more often. It's still common to see players deflecting their own shots (specifically backhanders) as well as the same shot being deflected 2-3 times with the first and last deflection coming from the same player, meaning that the puck would have to be doing it's best pinball impression. Not exactly realistic physics. Since part of why the amount of goals are still as high as they should be (more or less) is the complete lack of pass interceptions (and supersonic slapshots off of faceoffs, love those!), I decided to take a look at how common pass interceptions are, since there is no stat for this incredibly imporant part of both forechecking and defence, I had to rifle through a tonne of play-by-plays...AGAIN. I went through 14 whole games before I decided that looking through more at that point would just be a waste of time. In 14 games, a grand total of 1 pass interception was reported in the play-by-play, which means either the play-by-play is all kinds of messed up, or the match engine is. Only Riz knows which it is. At least I hope he knows which, because it's not like we can realistically figure it out since it's the only real source of output we have. So we can't exactly cross-examinate it with anything else to look for disparity.

[Shindigs]

Time to necro my own thread.

Because of this little entry in the patch notes for 1.4

In-Game Engine v80
=========================
- Updated ruling body rules usage
- Adjustments to deflection AI
- Small adjustments to scoring levels and scoring distribution
- Improvements to goalie AI when playing the puck
- Improvements to AI for selecting where to clear the puck
- Adjustments to coach linechanges near end of game
- Adjustments to passing AI
- Improvements to player pathing
- Fixes to some linechange commentary lines
- Fixed a rare divergency at the start of some overtimes
- Technical optimization changes to pathing to improve sim speed
- Technical optimization changes to collision detection to improve sim speed

So is the AI better at deflecting now? Is deflection not a dead attribute anymore? The answer may surprise you! (it won't)

Our subject is 50+ goal scorer Owen Tippett (across regular season+playoff), who looks a little bit like this:


Man that's what you want to see for deflections, good enough Bravery, 20 Strength, 20 Deflection, good enough Anticipation, kinda poor hand-eye (stickhandling). He's also on PP1 playing the correct side for his handedness down by the goal in the Diamond. If you can't find deflections from that spot, you can't find deflections. Oh, and his line is on Overload Slot in ES, so he should be in place to deflect a fair bit there too.

So what was the result? Also since his Slapshot is better than his Wrister, did he favor that?
Here are the types of goals in order of most to least, using the play-by-play classifications:
Snapshot: 20 (1 that deflected off 2 dmen and the goalie)
Wristshot: 11 (2 empty netters, 1 that deflected off a Dman and the goalie)
Backhand: 7
Non-descript shot: 5
One-timer: 4
N/A: 1
Squeeze past the goalie: 1
Deflection: 1
Slapshot: 0

Note that some goals are missing, 4 to be precise. I skimmed 96 games and obviously missed 4. Even if literally all of them were deflections it would still be an awful attribute, so really can't be bothered to recount all of them.

So not only did he not favor his Slapshot, he never scored with it. Unless the slapshot is implied on the one-timers, which would be fair enough.
The non-descript shot just said "shot the puck". So anyone's guess what kind those were. (anyone being someone willing to find that entry in events_eng.cfg)
The N/A just said Tippett scored, that was it. No shot/deflection event at all.
The single deflection goal happened like this:
Tippett snapshots at Crawford
The puck deflects off Crawford
The puck deflects off Tippett
GOAL!!!

So the only goal was a deflection of his own shot. Neat!

To figure out the starting point for using personal/team tactics rather than letting my coach do them I actually had to watch ~10 games in the 2D engine. And your players never seem to screen the goalie ever, that's why the shooting% of defencemen is always horrible as well (~3-5% is pretty standard). It's pretty hard to deflect if you won't go to the net, so whatever the AI adjustment was, it wasn't enough.

Finally, I'm a bit saddened that a player with 18 slapshot doesn't seem to ever use his slapper. I'm not surprised, but I was hoping that players favored their better shot. Doesn't seem like it though. And this guy shoots more than Ovi, so not like it's from a lack of total shots...