WEBVTT 00:00.900 --> 00:01.560 Hi there. 00:02.160 --> 00:09.320 In this lecture, I want to show you an example of an online opening database, there are several online 00:10.080 --> 00:16.710 one of my favorite sites I play on in the evenings, Lee Hastag, they have under the tools menu opening 00:16.710 --> 00:19.230 Explorer and this opening explorer. 00:19.680 --> 00:25.870 And you can see a large number of games for each move and it looks as though E4 is more popular. 00:25.890 --> 00:28.110 That doesn't mean it's necessarily the best. 00:28.610 --> 00:30.270 We, for example, look at the French defense. 00:30.660 --> 00:35.400 We see DeFore is the main move, followed by a free from the number of games being played. 00:36.420 --> 00:41.880 And you'll notice that actually the win rates are sometimes higher for the less played moves like the 00:41.880 --> 00:42.300 fray. 00:43.650 --> 00:48.320 So what we have here, Ashleigh, is an example of research methods. 00:48.330 --> 00:50.760 How do you actually research what you want to play? 00:51.060 --> 00:57.840 You might want to play a game that if we scroll down here, we see that the U.S. government, which 00:57.840 --> 01:02.370 is be free, actually has a 50 percent win rate for whites. 01:02.880 --> 01:03.970 So that's quite interesting. 01:04.040 --> 01:07.580 It seems a little bit more, in fact, than some of these others, 50 percent. 01:08.160 --> 01:13.220 So sometimes you want to do your own research and make sure that the sample size is decent. 01:14.130 --> 01:21.180 If it's, say, only two games something is based on, then it's not a significant you have to do your 01:21.180 --> 01:26.790 job when you do research in general is to remove bias, reduce bias as far as possible. 01:26.970 --> 01:30.250 You can quite often never remove bias totally. 01:30.750 --> 01:37.380 So, for example, if you're doing a research survey, you'd want as many people surveyed as possible 01:37.590 --> 01:43.080 and the diversity of those people surveyed to be as diverse as possible. 01:43.380 --> 01:50.970 For example, where they live different parts of a particular country or even globally, all different 01:51.300 --> 01:56.940 ages, gender, etc. You want to minimize bias generally. 01:57.930 --> 02:03.150 Otherwise you should definitely declare, you know, your scope of this research in a particular context. 02:03.510 --> 02:07.680 But the more examples you have, the more rare reassured you can have that. 02:07.680 --> 02:09.680 That's an interesting move to explore. 02:10.020 --> 02:14.610 Now, you know, you might want to explore, be free, because there is quite a large sample where 50 02:14.610 --> 02:16.640 percent when rape occurs. 02:17.070 --> 02:18.690 So factor this in. 02:18.690 --> 02:24.030 You know, sometimes you might see a fantastic percentage, but it is only, say, two games. 02:24.210 --> 02:25.140 It's pointless. 02:26.310 --> 02:28.740 That is not a very, very good sample size. 02:28.740 --> 02:34.530 So sometimes literally, you know, the number of sample size is evident if if it's like being used, 02:34.710 --> 02:41.160 the abuse of statistics to say something's got a 100 percent win rate, you know, that game that maybe 02:41.160 --> 02:48.660 two people won those two games, the opponents, you know, may have been off somewhere or been playing 02:48.660 --> 02:49.200 for a laugh. 02:49.440 --> 02:51.780 You know, it might have been a not very serious at all. 02:52.170 --> 02:55.470 So bias creeps in to the samples. 02:55.800 --> 03:01.510 And this is like a summary just statistically of those results. 03:01.510 --> 03:03.180 So you can use an opening explorer. 03:03.600 --> 03:05.670 And on this particular one, there's a filter. 03:05.670 --> 03:08.190 Actually, we can choose just the Masters. 03:08.190 --> 03:12.540 You see, we can actually change where the research is obtained from. 03:12.690 --> 03:21.000 So if we go to the Masters database and close that, you know, we see far less sample size and most 03:21.010 --> 03:27.330 the games means quite often one day games taken far more seriously than online games. 03:27.720 --> 03:33.510 So quite often you might want to do that switch, for example, to the Masters database to remove that 03:33.510 --> 03:37.590 bias of just frivolous online bullet games being included. 03:38.100 --> 03:45.660 So when you do research methods, think about the basics of reducing bias as far as possible. 03:46.050 --> 03:52.650 So general research methods indicates a larger sample size, a greater diversity within that sample 03:52.650 --> 03:53.220 size. 03:53.850 --> 03:57.810 And you generally need a skepticism to reduce bias more and more. 03:58.650 --> 04:02.490 And, you know, it's just a guide statistically to what's going on. 04:02.760 --> 04:06.420 If we look at our bee free here, it's still at five percent. 04:06.660 --> 04:08.580 So maybe, you know, be free is a good bet. 04:08.580 --> 04:14.310 If you want something offbeat off off the main line track, then be free. 04:14.310 --> 04:19.930 Could be the erratic gambling, which we do have a section for in this course, which often, you know, 04:19.950 --> 04:24.930 D5 Bishop Beita is played and they accept this pawn. 04:24.930 --> 04:25.680 And you play nicely. 04:25.680 --> 04:27.090 Fred, it's an interesting game here. 04:27.450 --> 04:30.180 You can see the statistics are quite interesting. 04:30.510 --> 04:32.340 So here's the draw percentage. 04:32.340 --> 04:33.120 Forty one percent. 04:33.390 --> 04:37.230 So that's still kind of favorable to why even in the Masters database. 04:37.500 --> 04:44.130 But even that is biased because, you know, what year was that based on after a certain year, you 04:44.130 --> 04:50.520 know, computer starting was starting to be used more and more and people perhaps found better ways 04:50.520 --> 04:51.540 of playing against that. 04:51.900 --> 04:55.020 So there's also a kind of time line bias. 04:55.020 --> 04:56.760 What years are we talking about? 04:57.930 --> 04:59.550 So just bear in mind when. 04:59.660 --> 05:07.810 Use an online database tool to research openings, so this this one is at least hastag tools, openings. 05:08.420 --> 05:09.170 But there's other ones. 05:09.170 --> 05:12.760 There's a chess gamescom that's also free to a certain extent. 05:13.220 --> 05:19.970 But when you get into a higher depth, you might need to pay back if you want one which is fully free. 05:19.970 --> 05:22.730 You know, Leech's dot org is one I'd recommend. 05:23.360 --> 05:27.760 And you can also watch be streaming in the evening quite often Kings Cross streams. 05:28.070 --> 05:30.590 So, yeah, these are very, very interesting tools. 05:30.590 --> 05:33.440 But just be aware of basic research methods. 05:33.440 --> 05:38.900 You might want to look that up and research the whole subjects of research methods and bias removal 05:38.900 --> 05:45.820 in general, because we're talking about his sample size from a certain population and a certain even, 05:45.860 --> 05:47.480 you know, time limit of the game. 05:47.720 --> 05:52.400 When we're switching to the Masters, we're talking about more serious one day chess, but we're also 05:52.400 --> 05:56.210 talking about a certain time in history as well. 05:56.750 --> 06:01.580 And if we had more and more filters, then we could say, well, what the you know, the grandmaster 06:02.000 --> 06:07.160 like games, you know, in recent years, you might want to filter by particular aspects. 06:07.580 --> 06:13.400 And if you want to get really, really professional about it later on, there is a local database you 06:13.400 --> 06:19.920 can get called caspase, which has been the kind of industry standard that chess professionals use. 06:20.330 --> 06:24.530 I certainly have benefited benefited greatly from chess base. 06:24.950 --> 06:27.870 It means also you can prepare for specific opponents. 06:27.870 --> 06:31.730 You can even prepare dossiers as though you're a real chess assassin. 06:31.940 --> 06:37.910 If you're playing someone the next day and one day you can prepare a dossier Hancey get an opening survey 06:37.910 --> 06:39.040 of their staff. 06:39.440 --> 06:41.420 So anyway, this is really quite interesting. 06:41.420 --> 06:47.270 There is there is a certain amount of filtering here for later, so you could stick to gyrated or whatever. 06:47.280 --> 06:49.250 So this helps reduce bias. 06:49.790 --> 06:56.300 So you'd hope that if you go to high ratings at least, or a longer time control, say Rappard instead 06:56.300 --> 07:03.860 of blends, that, you know, you'll get a certain amount of bias removal for even the online games 07:04.130 --> 07:06.170 to make them more sort of quantified. 07:06.530 --> 07:09.170 Or we could even go to like the classic all time control. 07:09.680 --> 07:17.180 And then we have another set of statistical research to have a look at by the key point I'm trying to 07:17.180 --> 07:17.720 make here. 07:17.930 --> 07:21.740 Yeah, be aware of the key priority of bias removal. 07:21.950 --> 07:23.330 What biases exist? 07:24.280 --> 07:29.530 And your particular data that you're looking at, so don't be carried away just because you see a massive 07:29.530 --> 07:29.990 win, right? 07:30.130 --> 07:31.990 It might be a low sample size. 07:32.230 --> 07:36.160 It might be just based on bullet charts or even worse, hyperbolic. 07:37.150 --> 07:40.510 It's all just about moving quickly, then just run the move generators. 07:40.870 --> 07:46.300 So would you really want to base your opening choices based on random moves in the extreme? 07:48.550 --> 07:55.120 So, no, no, you want to take control over your research methods and on this site you can do to a 07:55.120 --> 07:55.900 certain extent. 07:55.990 --> 08:00.670 But these filters, even if you just want to look at online trends, but if you want to look at the 08:00.670 --> 08:07.390 Masters, the over the board games where games are often taken, you know, one day at a time, one 08:07.390 --> 08:11.530 game is one day, then this Masters database is very, very interesting as well. 08:11.770 --> 08:16.030 And you see that, you know, the game is do get played if you look at the scene in the fans. 08:17.170 --> 08:18.850 So the main line moves are free. 08:18.850 --> 08:27.250 And if we follow this will find that the main line, the most statistically trodden line is the Neudorf 08:27.910 --> 08:28.630 A6. 08:28.900 --> 08:33.040 That's the Sicilian Neudorf and identifies it here as well. 08:34.330 --> 08:38.290 And that's definitely worth a try with Fisher and Kasparov played right off. 08:38.290 --> 08:44.050 We see the mainlines are often the most, you know, quality variations for both sides. 08:44.830 --> 08:49.810 But there is an interesting alternative which seems to score, you know, two percent higher, which 08:49.810 --> 08:55.300 is Queen The for trying to annoy the opponents with more positional trainmen with it be five. 08:55.510 --> 08:56.860 So we see all these stats. 08:57.850 --> 09:03.850 Just be aware of the biases that creep into the stats and you can take control to some extent of, you 09:03.850 --> 09:08.350 know, reducing those biases that are apparent in your windrow lost data. 09:09.730 --> 09:17.680 But also there is a case sometimes, you know, for finding novelty with engines, you know, checking 09:17.680 --> 09:22.270 with an engine, a particular position so you can even, like, turn on an engine and get an engine 09:22.270 --> 09:23.920 view of this position. 09:24.310 --> 09:28.330 And OK, things blats a tiny, fractional bit better. 09:28.720 --> 09:30.670 So you might add that into the mix. 09:31.240 --> 09:34.540 But the thing is, with engines, they have their own biases. 09:34.540 --> 09:38.470 When you look at an engine, the valuation assumes the opponent basically is an engine. 09:38.740 --> 09:41.260 Sometimes you just want easier to play stuff. 09:41.620 --> 09:47.650 And that's why in the over the board world, you know, even the UN sounds gambits which engines put 09:47.800 --> 09:52.570 down on engines are biased, of course, against unsound stuff. 09:53.110 --> 09:55.450 And in a way they influence opening fire. 09:55.450 --> 10:00.580 It's become more and more solid each year until, you know, things like the Kings injured the defense, 10:00.580 --> 10:04.780 which used to be very, very popular, or the Benoni, they listen. 10:04.780 --> 10:12.370 That sounds because the impact of engines, people doing a lot of analysis with engines and that also 10:12.370 --> 10:19.480 influences the data, what's been actually played and the results, and that affects the stats. 10:19.480 --> 10:22.930 And so it's an evolutionary process through time. 10:23.530 --> 10:29.110 So anyway, the key takeaway point here, I just wanted to emphasize take control of the biases. 10:29.110 --> 10:35.170 When you do research on openings with opening databases and you can quite often filter a little bit, 10:35.290 --> 10:36.940 especially how you can filter a little bit. 10:37.600 --> 10:41.440 And I would recommend checking the Masters as well as just the online. 10:41.590 --> 10:45.760 And if you can use the online chat time table and the average rating for sure. 10:46.000 --> 10:51.790 So that's a great way of reducing bias to make sure you're not just looking at statistical junk. 10:52.330 --> 10:54.160 OK, and so much.