THE SCA GRADING SYSTEM

Introduction Games against ungraded players
Example of a Scottish Grade Calculation - Previous player
Example Printout for Player X Ungraded Players
Printout Notes Worked example - ungraded player
Grading Formula Can you lose points by winning
Calculation - Graded Player Time Control
Drift Analysis Data Input
Minimum Sample Size Theory

Introduction

This article is derived from an article prepared by Douglas Bryson on the Scottish Grading System. It is included here as an example of an ELO based system. An article on the BCF Grading System is included elsewhere on this web-site. All programs for the SCA grading system were rewritten at the end of 19. This massive task was carried out by Willie Rutherford of Perth chess club who estimated he spent more than 300 hours in the rewriting and coding of the programs. Scottish players were invited at the time that this article first appeared to join Douglas in a vote of thanks to Willie for his excellent work.

To which perhaps we can add the thanks of chess players everywhere for the self-sacrificing efforts of those many nameless individuals all over the world who arrange and organise chess at every level for the rest of us.

Very many thanks to you all

Example of a Scottish Grade

Below is a sample of a player's grade worked out using the Scottish grading system, which is based on the ELO method. It is worth noting that this grading system is a comparative system, and individuals can have widely varying grades in different types of chess because of the pool of opponents they play. No system is perfect.

Top

Example printout Player X

Name:

Player X

 

Status:

Established Adult

Title: None

Main Club:

Auchtermuchty

(Member)

Previous Grade:

2405

Ranking: 4

The following games were played in Scottish events:-

Event

Op Num

Opponent

Status

Published
Grade

Adjusted Grade

Used
Grade

Expected
Score

Scottish Champ Jul

5190

Player 1

A

2465

2465

2465

0.416

Scottish Champ Jul

6599

Player 2

V

2415

2415

2415

0.486

Scottish Champ Jul

4858

Player 3

 

2370

2370

2370

0.549

Scottish Champ Jul

3332

Player 4

 

2380

2380

2380

0.535

Scottish Champ Jul

4168

Player 5

 

2245

2245

2245

0.711

Scottish Champ Jul

5400

Player 6

J

2360

2360

2360

0.563

Scottish Champ Jul

6171

Player 7

 

2280

2280

2280

0.669

Scottish Champ Jul

3157

Player 8

S

2105

2105

2105

0.851

Scottish Champ Jul

3881

Player 9

R

2205

2205

2205

0.757

Eligible Games in Scottish Events: 9 Points: 6

Top

Explanatory Notes

PUB GRADE

Grade as published in the most recent grading list.

ADJ GRADE

Where published grade is not 0, this is published grade plus any junior or newly graded adult edition applicable.

Where published grade is 0, this is estimated grade for opponent (if possible to estimate).

USED GRADE

This is ADJ GRADE limited to a maximum difference of 400 points from your own PREVIOUS grade.

TOTAL ELIGIBLE GAMES:

9

AVERAGE OPP GRADE:

2313.889

 

EXPECTED SCORE:

5.537

ACTUAL SCORE:

6.000

(66.7%)

Top

Grading Formula

A new grade for a previously graded player is given by the following formula:-

New = Old + 800 x (act. - exp.)
n

where:

new is the new grade

old is last season's published grade (2405)

act is the actual points scored (6.000)

exp is the expected points scored (5.537)

n is the higher of 30 and the number of games played. (30)

The resulting new grade for this performance is therefore 2417. (After rounding to nearest 5 the new grade published would be 2415).

Top

Calculation steps for graded players.

To calculate your new grade sum the actual points scored (1, ½, 0) for all your games in the season. Using table 1 (see below) sum the points you were expected to score. Your grade goes up when you score more than expected and down when you score less.

Newly graded adult

+25

Junior 10 and under

+80

Junior 11-14

+65

Junior 15-17

+50

Junior 18-19

+40

Junior 20

+25

Junior Age Unknown

+40

To obtain the expected score on each game calculate the gap between your grade and that of your opponent. For example a player of 1475 plays an opponent of 1660. The grading difference is 185, from table 1 the expected score for the lower graded player is 0.26 and the higher player 0.74.

Junior players and newly graded adults are assumed to improve their playing strength faster than established players- To compensate the opponents of juniors and newly graded adults a range of additions are added to the junior or new adult grades. These additions are shown in table 2.

e.g. If you play a junior aged 14 with a published grade of 1500 your expected score is calculated against a player rated 1565.

Top

Drift Analysis

To check the accuracy of junior additions a program calculates whether established adult players are scoring more than expected or less. If adults are scoring more than expected a small negative adjustment is made to all active players and if less than expected then a positive adjustment is made. This program attempts to achieve stability in grades from year to year.

Maximum 400 Gap

The grading gap between players is always calculated at a maximum of 400 points. A player with a published grade of 1600 plays a player with a published grade of 2200. For the 1600 player the opponent is classed as 2000, for the 2200 player the opponent is classed as 1800.

Top

Minimum sample size/grade

In the general grading formula the number of games is always set at the actual number played or 30 whichever is greater. The minimum published grade is set at 300.

Top

Games against ungraded players

If you play an ungraded player the game may count for your rating. Before the final grading list is calculated all the games of an ungraded player are collated. A temporary rating is calculated for the ungraded player based only on games against graded players. To be eligible for a temporary rating the ungraded player must have played a minimum 5 graded players and scored between 8-92%. The temporary rating is retrospectively slotted into the games records of all the opponents of the ungraded player (both rated and unrated opponents). The temporary rating will be listed in the detailed player print (see above) and will be close in value to the first published rating of the ungraded player. The reason why the temporary rating may be different from the final published rating is that the ungraded player may themselves gain additional games from temporary ratings of ungraded opponents.

This is a major improvement over the previous grading system where temporary grades counted for 1 event only.

Top

Calculations for a previously graded player

A player graded 1770. Actual points 55, expected points 50.338, number of games 84, drift -6.1.

New Grade = 1770 + (800 x (55 - 50.338))/84 - 6.1
= 1810 (rounded to nearest 5 points)

A player graded 2085, actual points 16, expected points 13.307, number of games 21 (set to 30), drift 21/30 x -6.1 = -4.27

New Grade = 2085 + (800 x (16 - 13.307))/30 - 4.27
= 2155

Top

Ungraded players

For an ungraded player to achieve a new grade he must play at least 8 gradeable games within 2 seasons. A gradeable game is one against an opponent with a published grade or an ungraded player with a temporary grade.

Calculation

Average score = total points scored divided by number of games played.

Average opposition = sum of opponent's grades divided by number of games played.

The average score is converted to a grading difference using table 1.

An initial estimate is made using the above data.

All the opponents grades are compared to the initial estimate to make sure the grades are no more than 400 points different from the initial estimate. This adjustment is necessary since a player could get a distorted first published grading by playing very high rated players and getting an inflated average opposition. Similarly with playing abnormally low opponents in relation to the player's true initial strength. If any opponents are more than 400 points different from estimate then make the grades exactly 400 points different. This will change the average opposition from 2 above.

The grading difference (positive or negative) is added to the average opposition. The temporary grade calculated is the value which all your graded opponents get in their player records.

If the ungraded player has played any ungraded players who gained a temporary grade from 6 above then these temporary grades will be added to the player record.

Using all the games against published grades and temporary grades calculate new average scores and average opposition.

All the opponents grades (initial published grades and temporary grades) are made to be no more than 400 points different from this second estimate. If any, grades have been changed by the 400 point limit then calculate a final average opposition otherwise use the average opposition from 8 above.

Calculate new grade from table 1.

Drift is applied and the grade rounded to the nearest 5 points.

Top

Worked example of a newly rated player

Opponent's age

Published grade

Adjusted Grade

Used Grade

Result

10

1270

1350

1336

0

12

1255

1320

1320

0

12

1125

1190

1190

1

13

1045

1110

1110

0

11

970

1035

1035

1

13

0

(1041)

1041

0

12

1170

1235

1235

0

14

1025

1090

1090

0

14

955

1020

1020

0

1l

655

720

720

1

14

0

(934)

934

0

A

1280

1280

1280

0

A

1130

1130

1130

0

A

0

(925)

925

1

The published ratings of "0" are unrated players. The adjusted ratings are the published ratings plus any junior additions. The ratings in brackets are temporary ratings.

Calculation

Calculate average score against players with a published rating only.
= 3/11 = 27.27%

Calculate average grade of opponents with a published grade (include any junior addition),
= (1350+1320+1190+1110+1035+1235+1090+1020+720+1280+1130) /11
= 1134.5

Average score as a grading difference in table 1 = -171.5

Initial estimate is 963.

None of the opponents used grades are more than 400 points different therefore no adjustment needed.

All the opponents get credit for a game against a 963.

The player in this example gains 3 temporary graded games.

Average opposition is 15380/14 = 1099, average score is 4/14 = 28.57%

Calculate estimate = 936. Since the game against the 1270 J10 = 1350 is more than 400 points above 936 then the 1350 grade is reduced to be 1336. Calculate new average opposition = 15366/14 = 1097.58.

-New grade = 934.6

New grade (rounded) = 935

Top

Can you ever lose points by winning?

The Scottish grading system is an averaging process over an entire season. The arithmetic of the grading formula means that in certain circumstances the marginal effect of winning against a player whom you have a very high expected score might result is a slightly lower positive change in your grade after that particular game than before.

For the above situation to arise both the following circumstances must apply.

    1. You must have played at least 3Ogames. (Less than 3Ogames and all wins are positive)
    2. You must be going up at least 64 points. (Less than 64 and all wins are positive)

For example,- actual score 40 points, expected score 30 points, 60 games played. The grading change on these figures is +133. You play another opponent against whom your expected score is 0.9 and you win. New figures are 41, 30.9 and 61, grading change is now +132.

However after 30 games have been played the value of any game is affected by the number of games you play. The Scottish grading system should be looked at as your total actual and total expected over the course of a season.

Top

Time Control

The fastest time limit for grading is all moves in one hour each per player. This requirement may be relaxed in junior chess.

Top

Data Input

The accuracy of grades produced is only as accurate as the quality of data input. All tournament organisers and clubs should use the standard grading forms available from the SCA. All events advertised in Scottish Chess will be sent grading forms. Grading forms are available from the chief grader and area graders.

Note that there is no longer any request for alphabetical. order for player names, grading order is just as acceptable.

There are still a number of clubs who do not send their club championships for grading. Make sure your club is a member of the SCA and gets all its internal events graded.

Top

Theoretical Background

The statistical basis for grading is detailed in the book "The Rating of Chessplayers Past and Present" by Arpad E Elo. (Now out of print - Editor)

 

Table 1. Expected Scores Or Average Performance

Diff

Exp.
+

Score

Diff

Exp.
+

Score

Diff

Exp.
+

Score

Diff

Exp.
+

Score

0

0.500

0.500

100

0.636

0.364

200

0.757

0.243

300

0.851

0.149

5

0.507

0.493

105

0.643

0.357

205

0.762

0.238

305

0.856

0.144

10

0.514

0.486

110

0.650

0.350

210

0.767

0.233

310

0.860

0.140

15

0.521

0.479

115

0.656

0.344

215

0.772

0.228

315

0.864

0.136

20

0.528

0.472

120

0.663

0.337

220

0.777

0.223

320

0.868

0.132

25

0.535

0.465

125

0.669

0.331

225

0.783

0.217

325

0.872

0.128

30

0.542

0.458

130

0.676

0.324

230

0.788

0.212

330

0.875

0.125

35

0.549

0.451

135

0.682

0.318

235

0.793

0.207

335

0.878

0.122

40

0.556

0.444

140

0.687

0.313

240

0.798

0.202

340

0.881

0.119

45

0.563

0.437

145

0.693

0.307

245

0.803

0.197

345

0.884

0.115

50

0.570

0.430

150

0.699

0.301

250

0.808

0.192

350

0.888

0.112

55

0.577

0.423

155

0.705

0.295

255

0.813

0.187

355

0.891

0.109

60

0.584

0.416

160

0.711

0.289

260

0.818

0.182

360

0.894

0.106

65

0.5

0.409

165

0.717

0.283

265

0.822

0.178

365

0.898

0.102

70

0.597

0.403

170

0.723

0.277

270

0.826

0.174

370

0.901

0.099

75

0.604

0.396

175

0.729

0.271

275

0.830

0.170

375

0.904

0.096

80

0.610

0.390

180

0.735

0.265

280

0.834

0.166

380

0.908

0.092

85

0.617

0.383

185

0.740

0.260

285

0.839

0.161

385

0.911

0.089

90

0.623

0.377

190

0.746

0.254

290

0.843

0.157

390

0.913

0.087

95

0.630

0.370

195

0.752

0.248

295

0.847

0.153

395

0.916

0.084

 

 

 

 

 

 

 

 

 

400

0.918

0.082

Note: After the difference in grades between two graded players has been calculated, the expected score of the higher graded player is found in the "+" column; that of the lower graded player in the "" column.

 

 

Start of article

Return to NCCA Home Page