EHM The Blue Line

JeffR wrote:As promised, here's the first in a series of more detailed looks at the game's internal systems. I'll be posting these as I write them up for the manual (where they'll get nicer formatting), but there's no reason not to post them now as well. I'm not going to give extremely specific numbers, e.g. the exact values for the chance of something occurring, but I may mention some general boundaries, e.g. what constitutes a normal rate of improvement for player attributes.

One word on terminology: I'm using the term "Attributes" here to describe the individual player ratings, e.g. faceoffs, acceleration, leadership, etc. I don't think that word actually shows up in the interface anywhere, but we use it internally to specify that level of editing within the broader category of "ratings," which can also include things like positional ratings, collective groupings of attributes, potentials, etc. The attributes are divided into six ratings groups: Offense, Defense, Goalie, Mental, Physical, and Hidden. The Hidden attributes don't change, so they won't be discussed here. Skaters use all of groups except Goalie; Goaltenders use only Goalie and Mental (and they replace a couple of attributes in the Mental group with goalie-specific ones.) You can see which attributes belong to which ratings groups on the "Ratings" screen of the Player Profile.

Offense, Defense, and Goaltending Attributes

The individual Offensive, Defensive, and Goaltending attributes each have a monthly chance at improving. The exact chance is the product of a two-step (three for some) procedure.

First, a check is made against the player's Coachability attribute; if this check is failed, the player's chance of improvement is penalized significantly - he just wasn't interested in listening to what the coaches had to say that month. The player's current Happiness level influences the check; an angry player with a bad attitude isn't going to see much of an improvement.

Next up is a test of the effectiveness of his team's coaches. The player is automatically assigned a "position coach", based on his preferred position: the coach on the team with the highest "Coaching Goalies/Defense/Forwards" attribute, as appropriate. (The automatic assignment eliminates some mindless busy work by doing what the player would do anyhow, assigning the best positional coach to each player.) If that coach's check is passed, the player's chance of improving the attribute receives a bonus; if it's failed, a penalty. The exact amount of the bonus or penalty depends on the player's training settings - if the attribute is part of the the primary training focus, it's a large bonus and small penalty; the secondary focus gets slightly less of a bonus and more of a penalty, and if it's not part of a focus at all, the bonus is the smallest possible and the penalty the largest.

And that requires a little more detail on the training focus system: the player's Offensive, Defensive, and Goaltending attributes are further separated into two subcategories, Mental and Skill:

Offensive Mental - Offensive Read, Getting Open, Screening
Offensive Skill - Passing, Puckhandling, Shooting Accuracy, Shooting Range
Defensive Mental - Defensive Read, Positioning, Shot Blocking
Defensive Skill - Faceoffs, Hitting, Pokechecking, Checking
Goalie Mental - Positioning, Poke Check, Rebound, Puckhandling
Goalie Skill - Blocker, Glove, Reflexes, Skating, Passing

(The "mental" attributes are ones that tend to improve as the player becomes smarter and more experienced, while the "skill" attributes are things that get better from simple practice, repetition, and technique refinement.)

Each player can be assigned one of those subcategories as his primary training focus and another one as the secondary. This is done by right-clicking the player's name on any roster screen (we had a separate training screen at one point, but removing that and building its features into the roster display made the user's workload a little easier.)

Finally, if the player is age 24 or under, they get a chance at a special "Prospect" bonus - this is a check done using the ratings of the coach with the highest "Coaching Prospects" rating on the team. When managing in a junior league, then, that rating should obviously be a strong criteria for hiring coaches.

Once all the bonuses and penalties are calculated, they're added to the base chance of improvement, and a simple pass/fail check is done against that adjusted chance of success.
If a player repeatedly passes every possible check, the bonuses will mean his chance of getting at least one point of improvement in that attribute over the course of a year should approach 100% (and it's possible, with some luck, that he could improve the attribute multiple times.) If he keeps failing every check, the chance of improvement will be near (but never equal to) zero.

These three attribute categories are all capped by separate potential ratings for each - once the maximum potential total is reached, the player won't improve further in that category. There is, however, a chance (and I won't specify exactly how or when) that those potential ratings can change. In general, most players shouldn't see their potential change by large amounts over time, but there is a chance that an exceptionally good or bad run of luck could move it up or down significantly.

Physical Attributes

The attributes in the "Physical" subcategory work a little differently. They're also checked individually, but only twice a year instead of monthly. The chance of improving them is entirely age-dependent, with the most rapid advances coming at a young age, and the odds gradually declining until the chance of physical improvement ends entirely at age 28. The player's best Physical Attributes are the most likely ones to improve further, as they get a bonus to their chance while the worst ones receive a penalty.

Mental Attributes

The Mental attributes are also handled differently. Some of them - Aggression, Determination, Bravery, and Coachability - have almost no chance of changing, just a low-odds check once a year that can cause a small (one-point) fluctuation in either direction. The other mental attributes (Team Player, Leadership, and Temperament) should increase slowly as a player ages - average luck will see a 1-point increase in each every 4 years - so veteran players will tend to have good numbers for those attributes. Mental attributes for goaltenders are a little different; for them, Mental Toughness and Stamina replace Bravery and Temperament. Both of those attributes belong to the group that sees steady improvement with age.

Neither physical nor mental attributes are restricted by potential ratings.

JeffR wrote:It seems logical to move on to this topic after talking about player development earlier. The player decline and retirement systems are intended to keep the game's player pool at a relatively steady level (after a few seasons of ramping up to full size); so far, they seem to be doing a reasonable job of this, so they're not likely to see any more major changes for the release.

Attribute Decline

The key point here: not every attribute gets worse in an aging player. In fact, (slightly) less than half of the player attributes will decline with time, primarily the physical ones: Acceleration, Agility, Balance, Fighting, Speed, Stamina, Strength, Getting Open, Passing, Hitting, and Checking. The actual decline mechanics are simple: 25% chance per attribute of a 1-point decline, checked twice a year once the player's decline period begins. At that rate, with typical luck a first-line NHL star will physically decline to NHL-average levels in about four years, and then to fourth-line levels in another three or four years.

Aging: the Attribute

Which brings us to the next part of the system: If you've played in Commissioner Mode, you may have noticed an attribute named "Aging" in the player's hidden ratings. The Aging attribute adds variation to player development by specifying when the player's decline starts. The low end of the scale can start a decline as early as the teens, while the higher (above 20) end can, in extreme cases, postpone it until the player is over 40.

Aging, however, is not a static attribute. It doesn't develop the way attributes typically do, but instead changes in response to certain events in a player's career. Most new players will start with a rating that leads to a decline beginning around age 25 - but if the player manages to stay employed and free of serious injuries, by the time he reaches his mid-20's his Aging number will likely have been pushed to a level where the decline doesn't start until age 30 or later.

The events that lower Aging are things that accelerate the approach of his decline - specifically, major injuries. Eventually, I'd like to tailor the changes to specific injury types, so concussion problems or major joint damage are more likely to cut a career short than, say, a broken hand. Right now, though, it's primarily tied to injury length.

I don't want to get too specific about events that improve Aging, so they can't be exploited, but in general they're things that you would associate with a player proving his durability. The general idea is, if a player retains enough skill and stays healthy, he has a chance of getting into a Howe- or Selanne-like situation where his level of play stays at a high level for an unusually long time.

Speaking of Gordie, I should note here that the historical game, with normal development turned off (the unchangeable default at the moment, we'll make it optional soon), recalculates player ability every year. So, Aging isn't a significant factor in historicsal play - that's all governed by the player's historical settings, which I'll discuss when I write in-depth about historical play.

Finally, note that the decline governed by Aging is independent of the attribute improvement I described earlier. So, if a player were to have a catastrophic series of injuries that sent him into a decline in his early 20's, the normal improvement that happens at that age would still be operating and counteracting the decline to some extent.

Retirement

Retirement comes in two forms: The first, more common at the higher levels of play, covers the veteran players who quit voluntarily at the end of a long career, knowing they can't play like they used to. The second scenario, much more common throughout most of the database, arrives when a player finds himself unemployed and can't or won't find another job.

The "unemployed retirements" are handled by the same system that governs all the player movement in non-playable leagues. A prolonged period of free agency will eventually lead to a retirement (although those should be uncommon except for the very worst players, the system is aggressive about relocating quality free agents to new teams.) And, in certain situations, players who have to move on from their current team might retire instead. That can cover a variety of cases: someone playing Canadian senior hockey or in a very low division in a European country may have nowhere else to go when his skills decline to the point where he's no longer good enough even for that level. Or, players may find themselves in a situation where a hockey career no longer makes sense - for example, there's a much higher attrition rate for the weaker players on Ivy League teams when they graduate, since using your Harvard degree to earn a living is a little more attractive than making $300 a week in the Federal League.

The "veteran retirements" take a number of factors into account. The player's current age, his Aging attribute, and the number of seasons remaining on his contract are weighed to determine a percentage chance that he might quit. Right now this retirement check is done for everyone on August 1; I'd like to make that a little more variable eventually, and include the possibility of mid-season retirements. Assuming typical attributes and short contracts, this should lead to retirements in the late 30's for most players who manage to establish a solid career in high-level leagues and remain free of serious injuries.

danzoeee wrote:quick question about player stats(pretty much).. Malkin has 3½ stars.. what the hell is up with that lol

DK_Hockey wrote:I dont like the player rating system in the game so far.
I have a game running with The Leaf´s and first season Phil Kessel was rated 8 and he was the best player all season with 88p finish 1overall in the leauge.
But next season he was only on rating 7. How is that possible?
I have so far played 4 season but I have not seen any of my plays getting better in the rating system.

If they could I would love to see a system like they have in EHM 2007 or like in Football manager.

JeffR wrote:Underpinning most of the AI changes is a new 10-point rating system, replacing the old OOTP-style five stars. Generally, players rated between 8 and 10 should form the core of your top two lines and defensive pairs, while those rated 6 to 8 fill the remaining roster spots, and those lower than 6 are, at best, useful as short-term replacements. As with the stars, the scale is always relative to the level you're currently at (if you're unemployed, it defaults to NHL-level.) So if you're playing with a team in the ECHL and look at an NHL roster, most of the players you see will be star-caliber by your league's standards (including, potentially, ratings higher than 10.0 - there's no cap on the top end of the scale.)

Keep in mind that it's not just a straight-line conversion of ratings to an arbitrary scale - to reach the 'first-line' levels of the scale players need skills that are relevant for the top lines (i.e., offense is weighted heavily), while getting a rating near the top of the ratings group associated with the bottom half of the roster requires more defensive skills and intangibles. That way, for example, a one-dimensional offensive forward who isn't good enough for the top 6 tends to drop right off the roster instead of getting a third- or fourth-line job - the "Keith Aucoin" effect. Finally, there's a special "goon bonus" for enforcer-type players that increases their ratings to a level where the AI will treat them realistically (the game lineup AI isn't doing that yet, though - that's coming as part of the game engine work this week.)

At the start of the game, most of the ratings you see will be fairly accurate representations of player skill. But newly-generated players can have significantly inflated or deflated apparent value. That's where your scouts come in: assigning them to a geographical area (particularly if they're from that area themselves) will reduce or eliminate the effects of inaccurate information about players in that area. (The geographical areas have also been significantly simplified.) I'll be elaborating on that in a systems discussion later this week.

JeffR wrote:Game Systems Discussion: Scouting

(I've had this written for a while, just hadn't posted it yet while busy with last-minute stuff, although some of you may have seen a shorter version in the newsletter. One more coming after this, about player generation and the background leagues.)

Explaining the new scouting system first requires a little bit of discussion about how scouting usually works in this type of game, and whether that reflects real hockey scouting.

The most common scouting system in sports strategy games that include one is discovery-oriented; Football Manager is probably the best example. You send out your scouts, and they report back if they've found any useful players (that's a big oversimplification of what can be done in FM, you can also target specific players to learn more about them, but discovery is the biggest part of the it.) You may get email/news notification if there's anyone that really stands out, and the information you know about players the scout "sees" is upgraded.

But that's not really appropriate for a modern hockey game. Unlike soccer, where it's possible for a player to develop in third-world obscurity and where they tend to be acquired by professional teams at younger ages, hockey is mainly played in a highly organized way in a small number of developed nations, and has mechanisms like drafts and minimum age rules. Draft-eligible players are highly visible and get watched and rated by many different professional and amateur scouts and organizations. There's a big demand for that information, and it gets disseminated widely over the internet. In short, it's very, very difficult and rare to find a hidden gem no one else knows about. Case in point: the big secret of the 2012 draft was Mark Jankowski, who was playing high school hockey in Ontario, not exactly a typical path to stardom. But he was ranked #74 in Central Scouting's North American list at the start of 2011-12, by midseason was getting media attention as a possible sleeper pick, and eventually wound up being Calgary's first-round selection. That's what passes for a "discovery" these days.

It doesn't make much sense, then, to build a scouting system for a hockey game with a discovery emphasis (and, honestly, those can get a little annoying, spamming you with messages about players that the scout has seen.) Instead, what we've done is to emphasize the quality of scouting in a way that separates the user's team from its competitors. In the new scouting system, young players get a "Consensus Opinion" attached to them when they're generated. This is the general opinion that scouts agree on when they're discussing players amongst each other, that tends to get published by the less-insightful scouting services and draft guides, and eventually gets settled on as conventional wisdom in public message boards. This can have either a positive or negative change in his perceived value by teams, overrating or underrating him in their eyes, by varying degrees (sometimes, the popular opinion is pretty accurate.) I won't say exactly how much it can change them, but it's enough to knock a legitimate first-round talent several rounds lower if he's badly underestimated.

Those inaccurate perceptions will gradually fade as the player ages, but not before the player advances past the early part of his career - i.e., the stage when teams have to make decisions about whether or not to draft or sign him while he's still developing. So, then, the role of scouting in this new system is to reduce that error - a good scout spending a lot of time in an area he's familar with can eliminate most of the consensus error even in cases where the popular scouting opinion about a player is very wrong. Your job is to put your scouts in places where they can do the most good.

When you start the game or take over a team, you can assign the scouts via the scouting menu - simply click on "add job" and you can choose a new target for him. Typically, you'll want to assign the scouts who are best at evaluating potential to amateur scouting roles, and those who are better at evuating current ability to scout professional players. It's also a good idea to place a scout in the region, typically near his home, where he is the most knowledgeable (we'll be adding a better indicator of whether or not he's in the "right" place to the scouting screen, right now it's not very helpful in making geographic choices.

The "ability" and "potential" ratings you see in-game are always post-scouting correction - your scouts' analysis is built into them (if you're unemployed, you just see the general opinion.) Likewise, the AI teams see players in the same way, dependent on the filter provided by their scouts - if they have few scouts or bad ones, they'll make more mistakes in evaluation.

You can change a scout's location whenever you want, but be aware that the game is tracking the amount of time your scouts spend in a given region and classification (pro/amateur). Constant coverage is the only way to optimize the error-reduction for a region; when determining how much to correct the consensus opinion, the game considers the amount of time your scouts have spent there, their scouting skills, and whether or not they "know" the area they're scouting. Near-perfect scores in all categories will eliminate almost all the error in the consensus opinion, but it'll be very rare to achieve that. Mainly, you're trying to minimize the effect, so when you go "off the board" with a draft pick that other teams would ignore, you've got legitimate reason (hopefully) to claim "he was a first-rounder on our list and couldn't believe he was still available."