testing data - database CA levels/distribution

[nino33]

In recent months I've been doing a lot of testing and posting the results; here's some looking at CA levels and distribution at startup, and for the first 10 years after startup (the data is taken from the day after the NHL Draft in June of each year).




Goaltenders CA, years 2018-2028 ingame

TBL 11.2

ingame Year	POS		100-200	150-200		120-139	140-169	170-200		total players
startup	G		321	26		100	48	2		53341
2019	G		346	28		109	47	3		40697
2020	G		337	26		103	51	5		47138
2021	G		336	33		105	51	5		50413
2022	G		341	30		109	52	5		50597
2023	G		333	32		105	55	6		50008
2024	G		338	35		105	54	7		49485
2025	G		340	34		101	53	8		47498
2026	G		344	36		98	54	6		46817
2027	G		339	34		97	51	6		48557
2028	G		316	20		99	40	5		51074

ECK 3.2

ingame Year	POS		100-200	150-200		120-139	140-169	170-200		total players
startup	G		307	32		57	32	12		96633
2019	G		314	29		47	27	11		88911
2020	G		309	28		49	24	11		91324
2021	G		320	28		56	22	11		94139
2022	G		313	26		47	29	11		94124
2023	G		322	28		48	29	10		94090
2024	G		338	29		35	37	8		94386
2025	G		349	27		25	40	9		92236
2026	G		356	31		28	38	9		91751
2027	G		340	31		23	44	6		93079
2028	G		340	33		28	42	7		93640

[nino33]

Defensemen CA, years 2018-2028 ingame


TBL 11.2

ingame Year	POS		100-200	150-200		120-139	140-169	170-200
startup	D		1029	56		331	98	4
2019	D		1035	70		318	113	8
2020	D		1013	78		300	128	8
2021	D		1037	80		303	126	10
2022	D		1037	77		311	123	9
2023	D		1041	83		301	126	9
2024	D		1028	87		300	129	9
2025	D		1019	86		292	131	9
2026	D		1015	79		283	134	6
2027	D		1005	78		280	131	8
2028	D		975	101		266	146	15

ECK 3.2

ingame Year	POS		100-200	150-200		120-139	140-169	170-200
startup	D		1642	111		293	137	20
2019	D		1618	120		307	137	36
2020	D		1495	126		275	139	49
2021	D		1435	128		244	162	50
2022	D		1385	142		203	188	48
2023	D		1368	155		199	201	47
2024	D		1318	171		188	206	49
2025	D		1290	174		181	208	48
2026	D		1284	173		167	206	49
2027	D		1265	178		151	205	50
2028	D		1217	181		150	202	49

[nino33]

Forwards CA, years 2018-2028 ingame


TBL 11.2

ingame Year	POS		100-200	150-200		120-139	140-169	170-200
startup	F		1836	107		613	155	9
2019	F		1784	124		561	180	19
2020	F		1754	133		508	216	24
2021	F		1777	148		495	236	26
2022	F		1784	163		479	245	25
2023	F		1790	168		477	249	27
2024	F		1811	169		481	250	26
2025	F		1818	173		487	248	30
2026	F		1816	179		491	249	34
2027	F		1832	188		506	242	37
2028	F		1883	173		503	246	37

ECK 3.2

ingame Year	POS		100-200	150-200		120-139	140-169	170-200
startup	F		1583	139		288	143	47
2019	F		1577	144		259	142	60
2020	F		1637	157		254	144	71
2021	F		1725	161		248	156	73
2022	F		1770	167		235	191	73
2023	F		1803	175		220	197	72
2024	F		1840	190		209	217	74
2025	F		1862	205		198	242	73
2026	F		1869	206		201	254	73
2027	F		1903	222		191	272	71
2028	F		1942	237		185	280	76

[nino33]

Taking a look at the 2018-2028 CA data...




• For goalies with a 100-200 CA, TBL averages 336 per year (range 316-336) and ECK averages 328 per year (range 307-356). The per year numbers increase a small amount for TBL; the ECK increase is more noticeable (the ECK database started with fewer).

• For defensemen with a 100-200 CA, ECK averages 1392 per year (range 1217-1642) and TBL averages 1021 per year (range 975-1041). I was surprised to see the ECK average per year drop significantly as years passed (25% less in 2028 than in 2018).

• For forwards with a 100-200 CA, TBL averages 1808 per year (range 1754-1883) and ECK averages 1774 per year (range 1577-1942). As the years pass ECK sees significant growth (23% more in 2028 compared to 2018).




• For goalies with a 150-200 CA, TBL averages 30 per year (range 20-36) and ECK averages 29 per year (range 26-33).

• For defensemen with a 150-200 CA, ECK averages 151 per year (range 111-181) and TBL averages 80 per year (range 56-101). When comparing TBL/ECK, there is a significant difference in the number of 150-200 CA defensemen, with ECK having close to double the amount. From 2018 to 2018 both ECK and TBL see significant growth in 150-200 CA defensemen.

• For forwards with a 150-200 CA, ECK averages 182 per year (range 139-237) and TBL averages 157 per year (range 107-188). Both databases see substantial increases in 150-200 CA forwards from 2018 to 2028.




• For goalies with a 170-200 CA, ECK averages 10 per year (range 6-12) and TBL averages 5 per year (range 2-8). I was surprised to see the ECK database see an almost 50% drop from 2018 to 2028 (TBL is at our above its average for all but startup and 2019, while ECK was below its average 2024-2028)

• For defensemen with a 170-200 CA, ECK averages 45 per year (range 20-50) and TBL averages 9 per year (range 4-15). Both databases start with far fewer than their 2018-2028 average; the ECK database has a lot more 170-200 CA defensemen per year than the TBL database (averaging 500% more).

• For forwards with a 170-200 CA, ECK averages 69 (range 47-76) per year and TBL averages 27 per year (range 9-37). The ECK database averages 255% more 170-200 CA forwards per year.

[nino33]

The above is looking at CA levels and distribution at startup, and for the first 10 years after startup.

The following data is from twenty to thirty years later: for the TBL database 2039-2049, and for the ECK database 2039-2048



Goaltenders CA, years 2039-2049 ingame

TBL 11.2

ingame Year	POS		100-200	150-200		120-139	140-169	170-200		total players
2039	G		320	36		82	53	5		52873
2040	G		337	35		78	54	5		52083
2041	G		340	34		74	53	6		52810
2042	G		331	32		86	55	6		54283
2043	G		321	35		87	55	7		54282
2044	G		329	34		82	57	7		54282
2045	G		323	31		77	53	7		54282
2046	G		314	29		78	50	7		54282
2047	G		316	31		84	47	6		54282
2048	G		320	28		83	47	7		54282
2049	G		319	30		83	46	7		54282

Goaltenders CA, years 2039-2048 ingame

ECK 3.2

ingame Year	POS		100-200	150-200		120-139	140-169	170-200		total players
2039	G		389	58		30	57	21		94752
2040	G		390	58		35	53	24		93184
2041	G		393	59		37	56	27		94172
2042	G		388	62		37	65	24		94901
2043	G		387	70		29	74	23		95311
2044	G		374	77		26	79	22		94903
2045	G		375	76		24	86	23		93772
2046	G		366	75		21	79	24		94620
2047	G		373	76		21	80	22		94674
2048	G		380	77		23	80	21		95393

[nino33]

Defensemen CA, years 2039-2049 ingame

TBL 11.2

ingame Year	POS		100-200	150-200		120-139	140-169	170-200
2039	D		1027	102		282	141	22
2040	D		1017	100		279	139	22
2041	D		1024	102		279	142	21
2042	D		1018	91		276	139	19
2043	D		1022	93		271	146	18
2044	D		1008	92		266	147	17
2045	D		997	91		280	144	17
2046	D		1014	91		269	148	16
2047	D		1014	93		272	150	17
2048	D		1001	93		267	146	19
2049	D		993	98		259	149	17

Defensemen CA, years 2039-2048 ingame

ECK 3.2

ingame Year	POS		100-200	150-200		120-139	140-169	170-200
2039	D		1237	173		77	192	61
2040	D		1246	175		82	203	55
2041	D		1235	194		80	195	69
2042	D		1218	198		69	215	59
2043	D		1217	193		67	213	62
2044	D		1190	204		62	207	67
2045	D		1183	200		54	215	69
2046	D		1199	206		49	214	75
2047	D		1196	214		46	207	76
2048	D		1203	219		52	213	74

[nino33]

Forwards CA, years 2039-2049 ingame

TBL 11.2

ingame Year	POS		100-200	150-200		120-139	140-169	170-200
2039	F		1834	159		507	227	39
2040	F		1828	164		515	232	38
2041	F		1818	159		515	231	39
2042	F		1837	173		508	234	39
2043	F		1843	164		507	236	35
2044	F		1850	166		510	241	35
2045	F		1867	164		500	249	35
2046	F		1859	171		504	254	36
2047	F		1856	173		488	265	33
2048	F		1865	180		498	263	32
2049	F		1874	176		506	260	35

Forwards CA, years 2039-2048 ingame

ECK 3.2

ingame Year	POS		100-200	150-200		120-139	140-169	170-200
2039	F		1926	284		170	319	92
2040	F		1917	291		156	315	95
2041	F		1930	287		148	319	91
2042	F		1948	281		135	324	94
2043	F		1962	282		123	332	95
2044	F		1975	277		112	353	93
2045	F		2000	291		118	342	95
2046	F		1993	291		118	350	93
2047	F		1988	290		120	355	93
2048	F		2002	289		107	362	94

[nino33]

Below I have copy/pasted my earlier summary of the 2018-2028 data in black font & added a 2039-2048 data summary in blue font.
The 2049 year testing data I have for TBL but not for ECK, so I used 2048 as the final year for both below.




Taking a look at the 2018-2028 CA data...
Taking a look at the 2039-2048 data...




• For goalies with a 100-200 CA, TBL averages 336 per year (range 316-336) and ECK averages 328 per year (range 307-356). The per year numbers increase a small amount for TBL; the ECK increase is more noticeable (the ECK database started with fewer).
• For goalies with a 100-200 CA, TBL averages 325 per year (range 314-340) and ECK averages 381 per year (range 366-393); TBL has seen a 3% reduction and ECK has seen a 14% increase.


• For defensemen with a 100-200 CA, ECK averages 1392 per year (range 1217-1642) and TBL averages 1021 per year (range 975-1041). I was surprised to see the ECK average per year drop significantly as years passed (25% less in 2028 than in 2018).
• For defensemen with a 100-200 CA, ECK averages 1212 per ear (range 1183-1246) and TBL averages 1014 per year (range 997-1027); TBL has essentially remained the same (1% reduction), and ECK has seen a 13% reduction (this is surprising).
• After being much higher in the initial years the total 100-200 CA defensemen for ECK in the final years of the 2018-2028 testing was 1284, 1265, and 1217 (similar to 2039-2048 range)...from startup it was two years of 1600s, two years of 1400s, three years of 1300s, and then seemingly settling into the 1200s. I didn't compile the data covering 2028-2039. I wonder if it stayed in the 1200s between 2028 and 2039? Or maybe during that time it worked its way up into the 1600s and then back down into the 1200s again?


• For forwards with a 100-200 CA, TBL averages 1808 per year (range 1754-1883) and ECK averages 1774 per year (range 1577-1942). As the years pass ECK sees significant growth (23% more in 2028 compared to 2018).
• For forwards with a 100-200 CA, ECK averages 1964 per year (range 1917-2002) and TBL averages 1846 (range 1818-1867). TBL has seen a 2% decrease, while ECK has seen a 10% increase.
• After starting much lower and growing substantially in the initial years, the total 100-200 CA forwards for ECK in the final years of the 2018-2028 testing was 1869, 1903, and 1942 (closer to the 2039-2048 range of 1917-2002, than the 2018-2018 average of 1774).

[nino33]

• For goalies with a 150-200 CA, TBL averages 30 per year (range 20-36) and ECK averages 29 per year (range 26-33).
• For goalies with a 150-200 CA, ECK averages 69 (range 58-77) and TBL averages 33 (range 28-36). TBL sees a 10% increase and ECK sees a huge 138% increase; previously the two databases had similar numbers of 150-200 CA goaltenders, and now ECK averages more than double TBL.


• For defensemen with a 150-200 CA, ECK averages 151 per year (range 111-181) and TBL averages 80 per year (range 56-101). When comparing TBL/ECK, there is a significant difference in the number of 150-200 CA defensemen, with ECK having close to double the amount. From 2018 to 2018 both ECK and TBL see significant growth in 150-200 CA defensemen.
• For defensemen with a 150-200 CA, ECK averages 198 per year (range 173-219) and TBL averages 95 per year (range 91-102); ECK has seen a 31% increase, while TBL has seen a 19% increase. There continues to be a significant difference between the databases in regards to average number of 150-200 CA defensemen, with ECK going from close to double to over double.


• For forwards with a 150-200 CA, ECK averages 182 per year (range 139-237) and TBL averages 157 per year (range 107-188). Both databases see substantial increases in 150-200 CA forwards from 2018 to 2028.
• For forwards with a 150-200 CA, ECK averages 286 per year (range 277-291) and TBL averages 167 per year (range 159-180). Between 2018-2018 both databases saw substantial increases in 150-200 CA forwards, and that pattern of increase has continued (in particular for the ECK database). TBL has seen a 6% increase while ECK has seen a 57% increase.

[nino33]

• For goalies with a 170-200 CA, ECK averages 10 per year (range 6-12) and TBL averages 5 per year (range 2-8). I was surprised to see the ECK database see an almost 50% drop from 2018 to 2028 (TBL is at our above its average for all but startup and 2019, while ECK was below its average 2024-2028)
• For goalies with a 170-200 CA, ECK averages 23 per year (range 21-27) and TBL averages 6 per year (range 5-7); TBL sees a 20% increase (though not much real change, going from averaging 5 to averaging 6). ECK sees a 130% increase in average per year (going from 10 to 23 per year).


• For defensemen with a 170-200 CA, ECK averages 45 per year (range 20-50) and TBL averages 9 per year (range 4-15). Both databases start with far fewer than their 2018-2028 average; the ECK database has a lot more 170-200 CA defensemen per year than the TBL database (averaging 500% more).
• For defensemen with a 170-200 CA, ECK averages 67 per year (range 55-76) and TBL averages 19 per year (range 5-7); ECK saw a 49% increase (from 45 to 67) and TBL more than doubled (from 9 to 19)
• For defensemen with a 190-200 CA, ECK averages 5 per year (range 3-7), TBL had none.


• For forwards with a 170-200 CA, ECK averages 69 (range 47-76) per year and TBL averages 27 per year (range 9-37). The ECK database averages 255% more 170-200 CA forwards per year.
• For forwards with a 170-200 CA, ECK averages 94 per year (range 91-95) and TBL averages 36 per year (range 32-39); ECK saw a 36% increase and TBL saw a 26% increase.
• For forwards with a 190-200 CA, ECK averages 11 per year (range 6-14) and TBL averages 3 per year (range 2-5).

[KevT90]

Hasn't there been multiple reports and complaints that the game could not generate high end CA players with its regen, when the pool of created prospects ran out ? Not sure if the ECK database used here is the one with the prospects generator, but even the TBL db appears to generate very similar numbers of high CA 30 years after its startup date.

[nino33]

Hasn't there been multiple reports and complaints that the game could not generate high end CA players with it's regen, when the pool of created prospects ran out ? Not sure if the ECK database used here is the one with the prospects generator, but even the TBL db appears to generate very similar numbers of high CA 30 years after its startup date.

Yes there have been such complaints (and this data doesn't really support such complaints).
And yes the ECK 3.2 database uses the prospect generator.




I think what people see is the AI using the full 1-20 range for Attributes for higher CA players (when human researchers/editors don't); it was discussed a bit here on July 16th https://www.ehmtheblueline.com/forums/v ... &start=525

This AI use of lower Attribute values is clearly visible here, where the "lowest to highest range" is given for each Attribute at startup and 30 years later https://www.ehmtheblueline.com/forums/v ... 10&t=20236 (interestingly, this data also shows the AI uses the 16-20 range more often than human researchers/editors)


I think some (a lot? almost all?) of the complaints are based on users seeing the AI produced players who have high CA/PA levels but will have significant flaws/weaknesses (very low Attribute values that human researchers don't use).....and if it's one of the Attributes that never changes in game or one that changes very little, then it'll never raise up to a reasonable level regardless of how much practice/training is given or how successful the player is ingame

Here's some examples of the Attribute range differences from the TBL database testing (looking at the top 372 NHL forwards by CA)

• at startup the range for Work Rate is 9-20, and thirty years later the AI range is 2-20
• at startup the range for Determination is 9-20, and thirty years later the AI range is 1-20
• at startup the range for Anticipation is 7-20, and thirty years later the AI range is 1-20
• other Attributes where the human researcher never went below 6 or 7, and the AI uses the full range (as low as a 1)...Loyalty, Pressure, Bravery, Flair, Important Matches, Pass Tendency and Vision

AND...I think the other thing that connects to people feeling this way is the AI will never produce NHL ready "star" teenagers.
The top CA range for 19 year olds for the AI is in the high 120s/low 130s.
For human researchers such players can be in the 150s/160s (which makes sense given how high the NHL average CA is...and as the game progresses this average CA only gets higher, which only makes the not ready for the NHL youngsters stand out more).




Another factor IMO...the old (pre 2015 EHM) ratings guidelines used 100-200 for the NHL CA range, as shown below, taken from here https://ehmtheblueline.com/wiki/index.p ... s_guidance
1st liner - 170
1st/2nd liner - 160
2nd liner - 150
2nd/3rd liner - 140
3rd liner- 130
3rd/4th liner - 120
4th liner - 110
4th liner/AHL - 100
AHL - 90
With such rating ranges a teenage high 120s/low 130s CA player (with a good number of really high Attributes already) maybe wouldn't seem as out of place given 100 CA is the low range

In TBL at startup the top 372 forwards have an average CA of 141.5 (with 126 being the lowest). With ECK it's a 148.7 average CA at startup (lowest 128). And thirty years later the average CA with TBL has risen to 146.8 (still 126 as lowest); with ECK it's risen 156.5 (lowest rises to 135).

Given that EHM used to see the NHL as using half the available CA/PA range (100-200), maybe the "better than 07 but still not a new game" EHM on Steam version might be struggling a bit with trying to have the NHL use less of the 100-200 range




And there's also another piece of the puzzle...it looks like the game (in an effort to address the initial lack of prospect talent complaints) dumps a lot of high PA players over just a few years in the early years of the game, and this really affects user perceptions.

From my my second post in this Reddit thread https://www.reddit.com/r/EHM/comments/c ... _about_pa/
Here's the 170-200 PA players in the TBL database (so roughly the 2023-2027 NHL Drafts), with all possible players added at startup
2005 YOB - 0 G, 4 D, 6 F
2006 YOB - 0 G, 1 D, 1 F
2007 YOB - 0 G, 1 D, 3 F
2008 YOB - 1 G, 1 D, 3 F
2009 YOB - 0 G, 1 D, 1 F

So a range of 2-10 170-200 PA players per year.
The four years previous years (YOB 2001-2004, the 2019-2022 NHL Drafts) had 20, 34, 65 and 60 players 170-200 PA players.
From an average of almost 50 170-200 PA players per year for the first half decade of drafts to an average of under 5 players with a 170-200 PA.

At startup, looking at all players age 19-33, the TBL database has only 58 170-200 PA players.
Should draft classes (YOB) have more 170-200 PA players than all NHLers age 19-33?

As noted in the Reddit thread, as it occurs with both the TBL and the ECK databases, I'm thinking the significant influx of high PA players for the initial drafts is likely game created (it's likely the game creating the years of too many 170-200 PA players).

[TurboJ]

Here's some examples of the Attribute range differences from the TBL database testing (looking at the top 372 NHL forwards by CA)
at startup the range for Work Rate is 9-20, and thirty years later the AI range is 2-20
at startup the range for Determination is 9-20, and thirty years later the AI range is 1-20
at startup the range for Anticipation is 7-20, and thirty years later the AI range is 1-20
other Attributes where the human researcher never went below 6 or 7, and the AI uses the full range (as low as a 1)...Loyalty, Pressure, Bravery, Flair, Important Matches, Pass Tendency and Vision

This. So good that you have gone through all this trouble testing the player pools developing. I have been reading this topic and I've been wondering how the player quality seems to dive after the first 4-5 years after seeing your data.

But what you say on this quote makes perfect sense. It's completely possible that many future star players get "hidden away" because they won't get drafted because of having such big, important weaknesses. Or they bust in the NHL because of them. If their initial stint in the NHL is particularly weak (because of having some of these very low important attributes), they may drop out of the NHL and stop developing because of the "star with lost motivation" effect if they play on an average league throughout their youth.

I've seen many decent prospects decline and even retire before reaching 27 years of age while having been 1st round candidates. Most of these players seem to either have very low atts on some important areas, or just be extremely one-dimensional. So the players might technically get to high CA, but they won't have useful atts and so they never really make it despite having "20" on many atts.

Example; I've seen players who have all green atts, many light greens, but then they have like 3 determination, or 4 work rate. Or similar level of anticipation, off the puck etc. And their performance typically equals that. Not to even mention the hidden stats which could theoretically have an even higher effect if disproportionetly bad.

One thing I seem to notice is that at least with the TBL rosters, there still aren't enough top level goalies generated, even with the data shown. Even if there are high-PA gaolies out there, somehow they don't end up being NHL level, or they bust badly while there. Very likely due to the effect you described.

I've been using the prospect generator with the TBL roster and IMHO it does make the drafts more realistic - at least more similar to the first years of gameplay. And the possibility of getting "top 5 drafted" players that are actually at their PA when 17 years of age adds to that realism, as that's exactly how real life draft busts happen - by means of the player not developing any further despite being super good for their age.

But again, thank you so much for all this data and your comments on them!!

[nino33]

Thanks much for the kind words! Much appreciated

[nino33]

I've been using the prospect generator with the TBL roster and IMHO it does make the drafts more realistic - at least more similar to the first years of gameplay. And the possibility of getting "top 5 drafted" players that are actually at their PA when 17 years of age adds to that realism, as that's exactly how real life draft busts happen - by means of the player not developing any further despite being super good for their age.

I know of the prospect generator but have never used it or looked into it at all.....does it give players a starting CA? Any Attributes? Or does the game AI determine starting CA and Attribute values?

I would think if a prospect generator is used to create players "right from startup" & Do Not Add Players is selected at startup, maybe the couple/few years of way to many high PA players could be avoided too.

Databases typically have 17 year olds pretty well covered (the players who will be drafted into the NHL at the end of the first season in play).
So I suspect "right from startup" would probably mean players age 16 and younger in most cases (ideally determine what's in the DB by Birth Year and then as needed adjust desired amounts of players at different PA levels for those first couple/few years)


P.S. With my 1974 DB project I'll be looking into this issue further myself, later this year I hope (I don't need the prospect generator HaHa but I do hope to avoid the problem of NHL level players having some "way to low" Attributes, and I do have some ideas on how to do so)

[TurboJ]

^^^
I really don't know the specifics of the prospect generator. It would be nice if I did. It does some things very well so with that knowledge we could enhance the future databases maybe. BTW, looking forward to your 1974 database! Do you plan to include draft prospects into the '80s with that DB?

[nino33]

BTW, looking forward to your 1974 database! Do you plan to include draft prospects into the '80s with that DB?

Yes! The plan is to have prospects right up to the current time

[Smirnov2Chistov]

Thank you again Nino for your data dissection!

I've always said that prospect development is always a double-edge sword. I've seen posters complain that regarding EHM07, prospect developing was always something that was 'too accelerated' and would be unrealistic. With the modern version of EHM, player development seems more streamlined and 'accurate' in my eyes. I tend to look at this game like real life modern hockey. There are drafts that produce high quality talent, and then there are drafts where prospects are not as strong in depth and quality.

In my saves, unless the player had unbelievable potential, I always had let them season a year or two within their respective junior leagues and then would make my decision from there. I'll have to look back at some prospects who were drafted first and see how well they have 'jumped'.

Nino, as always, huge thanks for everything you do!

[nino33]

Nino, as always, huge thanks for everything you do!

Thank you very, very much!


We share similar views on the game, and on how we play