version 9.10.10 testing

[jwr38]

I would just like to point out before all the things I am noticing that I like this program and I am only writing this as I want to make it better and as stable like version 1 of the sim. Since I created a player creator and looking to create it the best I can I am having a ton of test leagues. As I said in the addons forum area, I am creating presets so that others can create their own leagues quickly as easily with known outcomes. Right now I am finding it difficult to do. Oh just to add, I have been an official beta tester for hockey sims before. This isn't my first time.

I have set the league up so it has no morale, coach, injuries, suspensions, farm, the same schedule, same lines for each season. I leave all the defaults on so the leagues setting scoring, fighting, line strat, penalty mins etc are all out of the box defaults. I load this season for each test run to make my testing consistant (so no aging, just renaming the output file). I random the rosters after 10 seasons of testing and go again. I then create a new batch of players and try again. While not the prefect system it does show a lot of things.

- When printing out the html it states that there is a general error. I am not sure why, could it be because I have limited what it outputs as that makes it sim faster.

- forward ice time seems to be huge for forwards. My front line players with 80+ EN have an average ice time of 24-26 minutes each. This is on the default of 40 sec shifts. Of course this could be fixed by GM's but I put it out there.

- Fighting, I see that fighting is based off how physical your team is overall. The problem I have is that with the current setup you are unable to create 4th line goons as the fighting is based off ice time. So if I give a goon only 4-5 minutes a game he will only fight once or twice a season compared to players with 50 in fighting but have more than 14 minutes of ice time, he will fight way more often even if he has a high di. Of course a high fighting skill and high minutes means they go nuts in fighting. I find defensemen fight a lot more often than forwards as they are on the ice more often. My average for each team in fighting skill is around 45-50 and I have tried having goons but they just don't seem to like fighting .. I think all goon fighters have the Laraque syndrome in version 2.

- Defensemen with high DF and low PA and SC will put up more points and have better shoot % than guys with low DF and high PA and SC. I have players with 60's in PA and SC but high DF leading their team in scoring while D players with 80's in SC and PA but 50' in DF not even close.

-forwards with high PA seem to score more easily than SC players. I have even matched up a center with 80 in PA and 50 in SC with a 50 in PA and a 80 in SC. The high PA players seem to score more goals than the SC guys. I get league leaders with 65 in SC leading the league in goals scored while my high SC guys no where to be found.

- +/- this is completely random. One season I get players with +/- in the high 20's then with the same everything the next season they might have +/- of - 20 or more. I just can't explain why.

- fast skaters will always do better than slower skaters with strength even if they have te same stats else where. Players with strength of 60-70 but high skating generally do much better than a player with skating at 70's but strength of 80+.

- the best players often don't not lead the league in their respective areas. Most of the time it is lesser players leading the league. Best players rarely are the best players.

- If a team starts of badly they will continue to be bad even if their roster is vastly better than those above it. Remember I have no morale or coaches effecting this.

- teams, even top teams, can go from being the best team to the lowest team in the league when simming seasons. The players stats don;t seem to matter.

There are many many more little details I have found but these are the main ones. Without consistency I am able to create presets.

[dkrause71]

As someone who stated many of these same points in the other thread a few weeks ago. I do feel that is should be pointed out that Simon really made some nice changes to 9.10. The PP % which had been around 24% with goal slider at 50 was around 20% in the new version for me. I did see more of the high SC guys on the goal leader board. I did see some more variation in the attendance.

I did see the PP % go up when i lowered the goal slider still. I figure Simon has that on his checklist though. The variation in the attendance still could use some work. I had 4 teams at 95% and everyone else from 76-83%. The four teams were not even the top four in record- however three of the four were top five. And before anyone mentions star power as the difference- it wasn't a high star power team as the fourth. Also, the D score on the PP way too much.

I personally thought Simon made some great changes in the last update. If the PP could drop a tad more, the endurance for forwards could drop a tad more, and the high end abilities could be a little more pronounced (ie the passing guys shoot less, the high defensemen CK guys hit more, the high SC guys scored more)- i would be thrilled.


To recap the areas i would like Simon to consider tweaking more
1) Little more drop to PP %- with a focus on defensemen goal scoring on the PP. For example in my last sim- Housley had 25 goals for the season but 22 on the PP.

2) Little more tweaking of attendance averages. Maybe more weight to the star power.

3) Still would like to see the best teams play a little better. For example the best team in my file (95 Detroit) finished dead last in the West in my latest sim.

4) Get the passing players to shoot less which in turn would get them to score less. Basically, have the players have stats a little more in tune with the ratings.

5) Endurance for forwards could drop a tad more

I do understand people want some randomness. I do too. Just would like to see less. Say the goal would be one standard deviation away, right now we are at two away. Which is a big improvement from before.

If your next update is as good as this last one- i'll be very happy.

[starfrit]

I also notices many of these myself and I too think this should be looked at before relasing the gold version. Summer is a great time for us beta testers to help Simon make the final tweaks in the program so it can be ready 100% for september

[SimonT]

Quick and Dirty Answer :
Players with 99 PA and 99 of SC won't be your best scorer in your league because he'll pass as much that he shot.
Players with 50 PA and 99 of SC might be your best scorer in your league because he'll not pass and just shot.

[jwr38]

Quick and Dirty Answer :
Players with 99 PA and 99 of SC won't be your best scorer in your league because he'll pass as much that he shot.
Players with 50 PA and 99 of SC might be your best scorer in your league because he'll not pass and just shot.

So a good overall players (allrounder) will not produce the same as a player that can do one thing well? Why can't good players be good? Is this the same for defense? I have noticed that players with higher ck than df will check more than players with higher df than ck even if they are both high. (even if it is only by one point)

Like I said in my examples, I get players that are snipers next to playmakers and I find the playmaker scoring more goals than the snipers. Defensemen with high df rating and low pa and sc score more than players with pa and sc in the 80's and low df.

[SimonT]

My previous post is for goal only, not assist.

Players make decision on what to do with punk regarding a formula. As you can guest easily, SK is an important part of Skating Decision, PA is an important part of the Passing Decision and SC is an important part of Shooting Decision.

For a player with 99 SK, 99 PA, 99 SC, players have 33% chance of shooting (99 / (99+99+99) = 33%)
For a player with 99 SK, 99 PA, 60 SC, the players had 27% chance of shooting (60 / (99+99+60) = 23%)
For a player with 60 SK, 99 PA, 60 SC, the players had 27% chance of shooting (60 / (99+60+60) = 27%)
For a player with 60 SK, 60 PA, 60 SC, the players had 45% chance of shooting (60 / (60+60+60) = 33%)

P.S. It's NOT the formula but I'm simplifying so it’s easier to understand.

[dkrause71]

Quick and Dirty Answer :
Players with 99 PA and 99 of SC won't be your best scorer in your league because he'll pass as much that he shot.
Players with 50 PA and 99 of SC might be your best scorer in your league because he'll not pass and just shot.

I think most guys understand that.

But that doesn't explain stuff like this- Hawerchuk 88 PA, 74 sc, Hull 78 PA, 97 SC- on the same line, same minutes for the season (1705 to 1702). Hawerchuk ends up with 271 shots to Hull 219. So Hawerchuk scores 42 goals and Hull 29.

If this was a one time deal then i would say its just random luck. But i can site the exact same thing many times just in this one sim. Oates with more shots than Neely- same line and an even larger spread to PA/SC ratio for Oates (95,66). Neely played almost 200 more minutes than Oates and Oates still outshot him. Yet other example- same sim- Neal Broten has the lowest SC of any forward on his team- 39 goals with 80 PA, 60 SC ratings for New Jersey.

Like i stated before a few times. I would just like guys like Oates, Hawerchuk, etc to shoot less. Guys like Hull and Neely to shoot more. I get that to happen in v1 for the most part. I am not asking for the goals board to be a checklist of the top 10 SC guys. I still tend to see the good players out do the great players.

If i didn't want the game to work or if i didn't care, i wouldn't even bother to post. Simon i felt made some solid improvements the last update and i hope he continues work in these areas.

[jwr38]

OK time for a new test. This test removes any user inputted randomness. What I did was make a 20 players roster and then copied that same roster 16 times. So in other words every team has the same roster and lines. Each team will now play against itself basically. (look at the links below for the player rosters)



TESTING DETAILS START HERE
==========================
Now I took out coaches, morale, injuries, suspensions etc. Now team should just be acing off against itself. I expected to see consistency and I did! Some of it good, some of it not so
First of all the top scorers for each team was roughly the same. But the downside was the range! The same player could be +21 on one team while -28 on another. So that is a difference of 50 in the +/- stat. Shoot % had another big change. While looking through all the data I noticed that it wasn't randomness that was doing this.

I looked at the team standing. There I noticed a strong pattern. One team doing well, others average while a few doing very well. Then I looked in the schedule and saw an issue. It seems go into slumps without the moral effect. Teams will play badly, and the players stats go along with it. It is clearly not a random factor so there is a function that is making players play poorly.

To check if this was a long term thing I ran the league with 200 games. He it was starting to even out. While I still had a large spread it seems the slumping control was evening it out a little. Not much.

The more I look it seems to be some sort of team/line chemistry issue. Teams that are doing poorly all seem to have a few lines that don't work at all. No reason behind it is there? I ran this test a few times and it showed similar results each time.

I did this test as it should give pretty consistant results. By removing extra random information I hoped to see that players all score around the same values. With so much movement by teams/lines/players in their values but remaining consistant in the team scoring it demonstrates that the engine is over controlling the slumping process. The engine as such seems to be working fine.

With so much added to the slumping process now, what will happen when extra random factors are added in like; line changes, different rosters, strategies, morale, injuries, coaches etc etc. It seems that all the extra features don't have a impact as much as the slumping control. So a GM no matter the skill could still end up losing because his team was marked to slump that season.

Here is the links of the league for you to look yourself. (oh still getting a bug when outputting)

Normal small league size
http://wwihl.com/test/

200 game league
http://wwihl.com/test2/

[DHLB Commish]

Guys I simply must say: This is the best damned analysis thusfar. If this means another BETA versaion of the STHSv2 while this gets ironed out, I'm for it. Already people (GMs) noted issues as these in leagues I play in as well as the one I Commish and yet we always put it out to the "morale" effect or "'coach" effect or simple "random" effect. These tests seem to prove otherwise.

May as well iron this out as evely as possible now before it becomes a bigger issue later on.

Good job all!

[myRPHA.com]

This recent test and report back is by far the best yet. It is detailed and likely has the best possible variables to use in such a test. On one hand, I want a bit of randomness in my results (except for day to day scores) so that everything is not 100% predetermined. On the other hand, I agree that the engine seems fine but the effects of the slump are having a large effect on results.

As you may or may not know, my league is making the switch in a few months to STHS. We are jumping directly into v2 from FHL. I am a bit concerned over these results, but it's not make or break for us. Our league has, and has always had inflated ratings, which just adds another monkey wrench into the entire test phase. Our stars are not 75ov, they are 90ov fictional players, so it's hard to compare apples to oranges in my own personal situation. In the end though, I am hoping that the behavior of the program serves us well.

In testing with our own players and teams so far though, it's all over the map. Top tier teams can either win the cup or miss the playoffs from one season to the next. Keep in mind though, that GM intervention with lines, trades, etc may have a drastically different effect on results. Time will tell I guess as we are only about two months away from switching from the God awful FHL that, in 19 seasons, has provided little change to team performance. Even after thousands of trades in 12 calender years, the same teams finish high and low every season and God help you if FHL has your number.

As for James' testing, well done. I wish I could help more these days but my schedule is nothing short of pure insanity! FYI though, I am reading all of these posts every day and staying in the loop.

[SimonT]

P.S. On the streak and slump, people keep telling me that since V1. My answer is and will always be, other that the morale of team, its pure bad luck or good luck.

[jwr38]

Then your luck system is all screwed! What other reason can the same player with the same skills be a -28 while on another team he has +21? What is forcing the guy with -28 to continually play like crap while the +21 continually play well??

If it was down to just plain luck then the laws of averages would kick in and we would have closer results. Something is effecting it.

[SimonT]

Then your luck system is all screwed! What other reason can the same player with the same skills be a -28 while on another team he has +21? What is forcing the guy with -28 to continually play like crap while the +21 continually play well??

If it was down to just plain luck then the laws of averages would kick in and we would have closer results. Something is effecting it.

Difference between your last comment and telling me I’m incompetent? None for me.

[DHLB Commish]

Hi Simon,

I think your missing the point. Being defensive about it will not solve anything. The point of these analysis' is to help improve the already great program.

Currently while your calculations are sound there is an effect outside of the MORALE, COACHES etc that are causing certain game results which are simply outside of luck/randomness.

I am a very big follower of "statistics". Statistically, LUCK doesn'T exist, It is all about timing.

So Over an 80 game season, which is a pretty wide range, things should balance out. +/- 10 overall I would say would be reasonable. +/- 28 is going a little off. Over a 200 game season, they noted issues with slumps and such whats which are, in the end, something that should not be happening over such a long run. If it was pure 'luck" then slumps would not exist. A team would win or lose, 50/50, every match so, in the end, would end up over a long period of time with a .500 average of games won (over 80-200 games it would probably be a 60/40).

This is not what is happening.

I agree that having "the best team" win EVERY MATCH is not fun at all and useless in a sports game (anyone watch the canadiens last season? or the FIFA world cup right now?) but certain constants must remain in place which, currently, is not happening.

The tests they have done are pretty intensive and conclusive. Why not take a look at it?

The reaction you got that stated your formula was OFF was based on your own defensive reaction Simon. Everyone should just stand back a little and look at this objectively. In the end, we all want the same thing: A Great program.

If you read the comments and analysis, you will note that they did state that the MORALE and COACHES were set to OFF so would have NO effect on results... unless that is untrue tin the formula no matter what we CLICK in the settings in which case it is hidden codsing which you are not telling us about. Based on your comment, I am wondering if you read the analysis or simply pûshed it aside as if it was one of the V1 critiques you speak about. Please read them Simon. They are actually very well detailed and well done.

Maybe something was mis programmed? You aren'T perfect (none of us are). I mean, if you were perfect would we even have BETA versions of STHSv2? Would you not have simply taken out the STHSv2 as is and everyone would have said "what a genius!"?

We are all just trying tro help Simon. That is all. No one is attacking you or your program.

[jwr38]

Then your luck system is all screwed! What other reason can the same player with the same skills be a -28 while on another team he has +21? What is forcing the guy with -28 to continually play like crap while the +21 continually play well??

If it was down to just plain luck then the laws of averages would kick in and we would have closer results. Something is effecting it.

Difference between your last comment and telling me I’m incompetent? None for me.

sorry Simon I did not mean to come off rude. I guess i just felt a bit rubbed off on information I had collected as it does show clear patterns. If it was just a random setting I expect that I would get closer results without one team doing poorly for a long period while another performs much better than others. It doesn't happen at the start of the season but these slumps alway happen into the season. This is why I concluded that something is effecting the outcome of the game. I am trying my best to help not be a pain. People have brought up and issue and I was trying to help out to find what it could be.