!!! Bugfix-Release 2026-07-08 ONLINE !!!
New Database Update July 7, 2026:
More Reliable Data for Your Chess Training
Dear chess friends,
Before diving into the details of the new update, I want to take a moment to say thank you. The overwhelming positive feedback, the praise, and the incredible support from the community recently mean the world to me as the sole developer of this project. It is a huge motivation to keep improving the database for all of you!
This update introduces significant improvements that directly enhance the accuracy and structure of the game data. My main goal was to make the database even more reliable for your opening preparation and opponent analysis, featuring more precise player profiles and a much cleaner dataset.
Matching the Right Games to the Right Player
Nothing is more frustrating during opponent preparation than seeing games from players with similar names mixed up. To prevent this, I have heavily refined how player identities are handled:
- Locked Player Identities: Once a game is clearly linked to an official FIDE-ID, I now lock this connection within the system. Automated algorithms can no longer accidentally overwrite it due to name similarities. This ensures your opponent statistics remain completely reliable.
- Resolving Incomplete Names: Often, game notations miss first names (e.g., listing only “Aronian”). The system now intelligently cross-references these incomplete entries with historical performance data to assign them to the correct player profile. This means you won’t miss crucial games when preparing for your next match.
A Clean Database Without Annoying Duplicates
To keep your opening tree clean and ensure your statistics aren’t skewed by duplicate entries, I have completely overhauled the duplicate detection engine:
- Detecting Incomplete Games: Sometimes, a short, incomplete fragment of a game exists alongside the full match. The system now reliably identifies these using the first 10 moves and merges them, entirely independent of how the players’ names were spelled.
- Merging Name Variations: Different spellings (like “Fischer, Bobby” and “Fischer, R.”) are automatically standardized. This reduces duplicate entries caused by inconsistent tournament data by approximately 40%.
- Filtering out “Ghost Players”: The database now automatically cross-references player activity with rating histories. This fixes errors, removes non-existent “ghost players,” and makes the database roughly 10–15% more accurate.
Additional Practical Benefits for Your Training
- No Useless Mini-Games: All games with fewer than 11 moves have been completely removed. They hold no value for opening preparation and only unnecessarily bloat the database and your loading times.
- OTB and Online Strictly Separated: The data is now automatically and strictly categorized into “Online” and “Over-the-Board”. This allows me to provide two separate files, giving you precise control over which material you want to analyze.
- Verified Player Corrections: When players send me corrections for their own games, this data is now given the highest priority (identifiable in the database by the tag
[SourceQuality "-1"]), to ensure that these games appear in the database files. - Historical ELO Reference: To assist the system in correct player matching and to process missing rating data more precisely, I now use the ELO history file provided with the program Scid vs PC as a reliable reference.
Despite all these optimizations, I want to be transparent: achieving a 100% accurate automated process is factually impossible given the massive raw dataset of over 35 million games. However, I have implemented rigorous automated validation checks to ensure the highest possible standard of data quality is maintained for you. If you have any questions about the new structure or how to integrate it into your chess software, please feel free to reach out to me!
The curated chess archive for your preparation
End the PGN chaos from the internet. Lumbra’s Gigabase filters millions of raw data through a highly efficient Rust pipeline into a cleaned, high-performance archive for Scid vs. PC/MAC and Scid 5.x.
The databases at a glance
- OTB Database: Over 10.3 million over-the-board games from more than 900,000 players.
- Online Database: Over 7.2 million curated games (Elo 1800+) from more than 125,000 players.
Current Version: July 2026
Databases: 2026-07-08
Incremental update: none
Time-range OTB PGN files: 2026-07-08
Opening books time ranges: 2025-02-06
Opening books ECO: 2025-02-06
Opening books NIC codes: 2026-05-05
Highest data quality through Rust deduplication
Lumbra’s Gigabase is characterized by an uncompromising focus on data quality. My custom-developed system written in Rust identifies exact and subsumed duplicates fully automatically. Short games (under 10 half-moves) are filtered out just as consistently as flawed name strings.
Partner Showcase (TBD)
This section is dedicated to project partnerships, chess training apps, and integrated platforms. Some of the links in this Partner Showcase are affiliate links. If you purchase software or subscribe through one of these links, I’ll receive a small commission to support this project—of course, the price remains exactly the same for you.
Chessever
Chessever is the mobile chess streaming platform. Follow live tournaments, keep an eye on grandmasters, and analyze games in real time on iOS and Android. If you subscribe within 14 days of registering, I’ll receive a small commission at no extra cost to you.
Thank you for your support!
Monthly Updates
The database is updated regularly on the first Tuesday of every month. To stay up to date, it is sufficient to import the monthly update file in PGN format into your existing Scid database. A complete re-download is not strictly necessary.
Support for this project
Server operations, the Rust infrastructure, and the continuous curation of the data incur monthly costs. If this free archive helps you with your chess preparation, I would appreciate a contribution via Buy Me A Coffee.
8 responses to “Free chess game database for Scid”
-
OK, I have played through the game and figured it out. I should have parsed the moves as follows:
1. d4 d5
2. c4 e6
3. Nc3 c5
4. cxd5 exd5
5. Nf3 Nc6
6. g3 Nf6
7. Bg2 Be7
8. O-O O-O
9. Bg5 Be6
10. dxc5 Bxc5
11. Ne1 d4
12. Bxf6 Qxf6
13. Ne4 Qe7
14. Nxc5 Qxc5
15. Nd3 Qb6
16. Nf4 $5 Rad8 $5 ( 16… Qxb2 $6 17. Nxe6 fxe6 18. Rb1 { favours White (Rubinstein – Tarrasch, Karlsbad 1923). } ) ( 18. Rfe8 { stems from Flohr – Euwe, Amsterdam 1932). } ) 18…
Qd2 19. Rfe8
Rfe1 20. h6
b3 21. Nb4
Rac1 22. a5 $5
Nxe6 23. Rxe6
Bh3 24. Rf6
Rc8 25. Rxc8
Bxc8 26. g6 ( 26. d3 $2 { Not } exd3 27. Nxd3 $2 { due to } Re8+ 28. Kh7 Qxd3+ )
29. Rd1 Rd6
30. a3 Nc6
31. Qxh6 $5 Qxb3
32. Rc1 $2 ( 32. Qc1 $1 { Prudent is } ) 32… Rd8
33. Bf5 $1 Qxa3 $1 ( 33… gxf5 $2 34. Rxc6 $1 { draws. } ) 34…
h4 35. Qf8
Qd2 $1 36. Qe7 ( 36. gxf5 $2 Rxc6 $1 37. bxc6 Qg5+ { draws again. } )
38. Bd3 $1 Qe5
39. Ba6 Qe7
40. Qg5 Qxg5 $2 ( 40… Rb8 $1 41. Qxe7 $6 Nxe7 42. Rc7 Nd5 43. Rxb7 Rxb7 44. Bxb7 Nb4 $1 { wins. } )
45. hxg5 d3 $1
46. Bxd3 $1 ( 46. exd3 $2 { Not } Nd4 ) 47. a4 $1
Ra1 48. Ra8
Be4 49. Nd8 $1
Kf1 50. Ra7
Bd5 51. Kf8
Rb1 $1 52. Ke7
e4 53. Kd6
Ke2 54. Kc5
Rc1+ 55. Kb4
Rb1+ 56. Kc3
Rc1+ 57. Kb2
Rc4 ( 57… Kd2 $1 { Perpetual check brings } 58. b5 $1 Rc5 $1 59. b4 Rc2+ 60. Kb1 Rc1+ ) 61. Ka3
Rc7 62. Ra8 $1
Re7 63. Rb8 $1
Re8 64. Kb2
Bxf7 65. a3
Rg8 $2 { White misses the last escape. } ( 65… f4 { Right is } 66. Nxf7 $1 Rxb8 67. Nd6 f5 68. a2 fxg6 $1 69. a1=Q g7 70. Qa6+ Ke1 $1 71. Qa5+ Kf1 $1 72. Qb5+ Kg1 $1 73. Qc5+ Kh1 $1 ) 74. b5 $1
f4 75. b4
f5 76. b3
Bxb3 77. Kxb3
Rxg6 78. Nf7 $1 { Dyckhoff showed his great fighting spirit again. } ( 78. Nf7 { The pointe is } Ra6 79. Nxg5 )
0-1The parentheses and braces are correctly placed, as long as one ignores a few extra close parentheses, but the move numbers are incorrect in both placement ( the number should precede the white move) and value (eg move 29. is actually move 25.) This is actually a very interesting game, I would hate to exclude it. I have only looked at the actual moves, I have not asked if the comments and annotations make sense.
-
Hello, I have downloaded your 4/12/2026 OTB pgn files and I am attempting to parse them with a finite state machine. I believe I successfully parsed the files in your 7/1/2025 database, but now I am having problems with matching ( and ), [ and ], { and }. Let me give an example. From the 1900-1949 pgn file, line 568203. I have manually separated it into one actual move per line as
1. d4 d5
2. c4 e6
3. Nc3 c5
4. cxd5 exd5
5. Nf3 Nc6
6. g3 Nf6
7. Bg2 Be7
8. O-O O-O
9. Bg5 Be6
10. dxc5 Bxc5
11. Ne1 d4
12. Bxf6 Qxf6
13. Ne4 Qe7
14. Nxc5 Qxc5
15. Nd3 Qb6
16. Nf4 $5 Rad8 $5 ( 16… Qxb2 $6
17. Nxe6 fxe6
18. Rb1 { favours White (Rubinstein – Tarrasch, Karlsbad 1923). } ) ( 18. Rfe8 { stems from Flohr – Euwe, Amsterdam 1932). } ) 18… Qd2Here the open parenthesis ‘(‘ in move 16 is not matched until move 18. My code believed the alternative move sequence included “17. Nxe6 fxe6 18. Rb1”, but this can’t be because there is no other move 17 in the entire game. So my question is, how might my program know when a comment or list of alternative moves is ended?
Every ‘(‘ appears to be matched with a ‘)’, but there is often an actual move enclosed between them. Examples below are moves 26-28, 33-34, 36-37, 40-44, 57-60, and 65-73.
19. Rfe8 Rfe1
20. h6 b3
21. Nb4 Rac1
22. a5 $5 Nxe6
23. Rxe6 Bh3
24. Rf6 Rc8
25. Rxc8 Bxc8
26. g6 ( 26. d3 $2 { Not } exd3
27. Nxd3 $2 { due to } Re8+
28. Kh7 Qxd3+ )
29. Rd1 Rd6
30. a3 Nc6
31. Qxh6 $5 Qxb3
32. Rc1 $2 ( 32. Qc1 $1 { Prudent is } ) 32… Rd8
33. Bf5 $1 Qxa3 $1 ( 33… gxf5 $2
34. Rxc6 $1 { draws. } ) 34… h4
35. Qf8 Qd2 $1
36. Qe7 ( 36. gxf5 $2 Rxc6 $1
37. bxc6 Qg5+ { draws again. } )
38. Bd3 $1 Qe5
39. Ba6 Qe7
40. Qg5 Qxg5 $2 ( 40… Rb8 $1
41. Qxe7 $6 Nxe7
42. Rc7 Nd5
43. Rxb7 Rxb7
44. Bxb7 Nb4 $1 { wins. } )
45. hxg5 d3 $1
46. Bxd3 $1 ( 46. exd3 $2 { Not } Nd4 )
47. a4 $1 Ra1
48. Ra8 Be4
49. Nd8 $1 Kf1
50. Ra7 Bd5
51. Kf8 Rb1 $1
52. Ke7 e4
53. Kd6 Ke2
54. Kc5 Rc1+
55. Kb4 Rb1+
56. Kc3 Rc1+
57. Kb2 Rc4 ( 57… Kd2 $1 { Perpetual check brings }
58. b5 $1 Rc5 $1
59. b4 Rc2+
60. Kb1 Rc1+ )
61. Ka3 Rc7
62. Ra8 $1 Re7
63. Rb8 $1 Re8
64. Kb2 Bxf7
65. a3 Rg8 $2 { White misses the last escape. } ( 65… f4 { Right is }
66. Nxf7 $1 Rxb8
67. Nd6 f5
68. a2 fxg6 $1
69. a1=Q g7
70. Qa6+ Ke1 $1
71. Qa5+ Kf1 $1
72. Qb5+ Kg1 $1
73. Qc5+ Kh1 $1 )
74. b5 $1 f4
75. b4 f5
76. b3 Bxb3
77. Kxb3 Rxg6
78. Nf7 $1 { Dyckhoff showed his great fighting spirit again. } ( 78. Nf7 { The pointe is } Ra6 79. Nxg5 )
0-1Thanks for any assistance.
-
Hi Team, not sure if this has been addressed yet, but 2025 games are mostly cropped, many games have moves ending in around 8 to 10 moves.
-
-
Best database ever. It would be really nice to release a lite version with only the important games above 2200 elo.
-
Thank you so much for all your hard work. I have been using SCID for many years, but now have a great database to go with it. I’m going to share this with all my students. Now, to work out how to buy you a coffee :-)
-
If this page is an chess move. It will be a brilliant move.
As my favorite Gm would say.
Sanx for ze games. – Daniil Dubov -
This is a fantastic resource! Thank you so much for making this collection.
I’m in the process of writing a book on romantic style chess openings, and your database has been very helpful in finding games from some of the old masters.
Leave a Reply