Tuesday, 17 September 2019

Does EA Sports Use A Formula?

If there is one hotly-anticipated game that's released each year, it has to be EA Sports' NHL game. While NFL fans usually salivate over the latest Madden entry, the NHL games have been among the top-selling games each and every year. It's fun to see the hot rookies introduced into the game, and the new features always seem to grab the attention of gamers. However, I always struggled to understand how EA Sports came up with their player ratings each year because it seems like they kind of make it up rather than having some hard-and-fast formula on which they rely.

I might be entirely wrong, and I assume I am, but how does EA Sports differentiate a player rated 100 overall compared to a player rated 98 or 99 overall? How do they determine year from year what player ratings should be if they aren't factoring in real, measurable stats from the previous season? What is the formula they are using to determine these ratings?

I wanted to know this answer after seeing the work done on Hockey-Graphs regarding player ratings from NHL '93 to NHL 2005. As seen on the image below, Hockey-Graphs did a comparison of ratings given in 1993 versus the ratings that EA Sports handed out in 2005, and there's a pretty clear distinction between the two games.
As you can see, the distribution of ratings in NHL '93 was much more even in terms of the entire collection of players whereas NHL 2005 seems to have a very large amount (half of all players) rated between 71-80. In fact, if you look at the graph on the site linked above, there's no player rated 40 or lower in any NHL game after NHL '96.

Is EA Sports artificially boosting player ratings? Possibly. Some of that, as Hockey-Graphs also hypothesizes, is likely due to the NHLPA agreeing to have its players represented in the game. It's very likely that none of the players appreciated being rated particularly low, so I imagine that EA Sports cut a deal with the NHLPA to have most players rated 70 or better. It's easier to keep everyone happy that way.

It's here, however, where we start to see some of the cracks in this deal as there are no fewer than four players in NHL '98 rated 100 - Dominik Hasek, Sergei Fedorov, Eric Lindros, and Jaromir Jagr - while there were no fewer than three players rated 99 overall - Ray Bourque, Patrick Roy, and Pavel Bure.

Hasek won the Hart and Vezina Trophies in 1997, so it's understandable to see why he'd be rated as the best goaltender in the game. Jagr finished sixth-overall in scoring, Lindros was 23rd-overall in scoring despite playing just 52 games, and Fedorov was 43rd-overall in scoring behind snipers such as Travis Green, Andrew Cassels, and Josef Stumpel.

Personally, it's hard to rank a guy like Fedorov in NHL '98 as a 100 overall if he's third on his team in scoring, some 22 points behind second-place and 24 points behind the leader. It's even harder to stomach when the guy who was second in scoring on the Red Wings - Steve Yzerman - was rated as a 76 in NHL '98. Yes, Fedorov did lead the Red Wings in scoring in the playoffs, but it's not like Yzerman fell off the map in helping the Red Wings to a Stanley Cup in 1997. How did EA Sports come to these player ratings?

To give you an idea how much the needle moved with regards to artificially improving player ratings, consider the plight of goaltender Bob Essensa. Essensa was mostly a journeyman goaltender who found success in Winnipeg before being discarded by the Red Wings following a trade. He bounced around as a backup netminder for a number of years, finding NHL work with Edmonton, Phoenix, Vancouver, and Buffalo.

In NHL '95, Essensa was a 51 overall with the Winnipeg Jets. In NHL '96 and NHL '97, he didn't even make the game before resurfacing in Edmonton in NHL '98 as a 58 overall. He followed that up in NHL '99 with a 56 overall before jumping up to 75 overall in NHL 2000. How on earth did Essensa make a 19-point overall improvement on his rating as a backup netminder with the Oilers?

Here are those three seasons of stats compiled by Bob Essensa.
In 1997-98, Essensa actually posted better overall numbers as he went 6-6-1 with a 2.55 GAA and a .913 save percentage which EA Sports determined would lead to a 56 overall rating, two points worse than one season earlier! And in 1998-99, Essensa went 12-14-6 with a 2.75 GAA and a .901 save percentage, but was rated a 75 in NHL 2000?!? Clearly, something is way off with Essensa's rating in 2000, but maybe it was just a one-off?

The issue with this is that it wasn't a one-off. Essensa's overall player ratings from 2001 until 2003 were 72, 76, and 70, respectively. Essensa's final rating of 70 was based on a season in which he went 0-5-0 in nine games for Buffalo while posting a 2.91 GAA and an .850 save percentage! How is he rated 70 for that season, but a 56 in NHL '99 when he statistically was superior in every category?

It seems that trophy winners received significant bumps in their ratings. After winning the Hart Trophy in 1994, Sergei Fedorov's rating jumped from 84 in NHL' 94 to 99 in NHL '95. After winning the Hart Trophy in 1994-95, Eric Lindros jumped from an 84 overall rating in NHL '95 to a 100 rating in NHL '96. Mario Lemieux was rated as a 100 overall in NHL '94 after winning the Art Ross Trophy in 1993. Jaromir Jagr went from an 87 in NHL '95 to a 97 in NHL '96 after capturing the scoring title.

Ironically, Nicklas Lidstrom won his first Norris Trophy in 2000-01 and the Conn Smythe Trophy as MVP for the playoffs, and saw his player rating go from a 94 in NHL 2001 to an 84 in NHL 2002, the lowest he was rated by EA Sports since NHL '97. It seems, as far as I can tell, that his drop of two points in scoring between 1999-2000 when he posted 73 points and the 71 points he scored in 2000-01 was the impetus for the drop in his player rating of ten points. Yet when he scored just 59 points in 2001-02 and won his second Norris Trophy as the league's bets defender, Lidstrom received a 98 overall rating in NHL 2003.

Simply baffling.

So what does it all mean? It's very likely that EA Sports is simply another marketing tool for the NHL and the NHLPA. Players are given favorable ratings thanks to the number of players who play video games in today's NHL, and having half the NHL rated 71-80 as Hockey-Graphs showed means that there are very few players who fall into the "not very good" category that early NHL games had with their player ratings.

Of course, I could be wrong and perhaps there is some algorithm or advanced mathematical formula that gives EA Sports precise player ratings based on the individual skills and abilities that are rated for each player. If EA Sports does indeed have this formula, I could see them wanting to ensure it remains under lock and key so that others can't replicate their advanced math and build a player rating system of one's own.

In seeing the artificially-boosted player ratings as shown in the Hockey-Graphs images, I thought of a key line from The Incredibles uttered by Syndrome as he planned to market his technology as a "superpower" so everyone could be a superhero. He said,
While parity may be real in the NHL today, it shouldn't be among player ratings in the NHL series. The is simply no way that a 59-point season for Jeremy Roenick is the same as a 79-point season for Alexander Mogilny, yet both were rated as a 90 overall in NHL 2004 after posting those point totals in 2002-03. Frankly, that's just downright ridiculous.

Until next time, keep your sticks on the ice!

No comments: