WEBVTT 00:00.630 --> 00:05.940 In this video you will learn how to create expanding windows so you will see in a minute what I mean 00:05.940 --> 00:06.360 here. 00:06.360 --> 00:14.550 And first of all we import pandas and we are working with the S&P 500 data set and we are only interested 00:14.550 --> 00:16.600 in the last 10 years. 00:16.740 --> 00:20.040 So this is actually nothing new here. 00:20.040 --> 00:25.440 And in the last videos we created rolling windows with a constant window size. 00:25.440 --> 00:32.720 So for example 10 days and by moving forward one day or one timestamp we actually drop the oldest timestamp 00:32.730 --> 00:36.140 and at the most recent time stamp. 00:36.150 --> 00:43.440 So for example we could calculate the rolling mean or the simple moving average for our clothes column 00:43.440 --> 00:50.890 here and our fixed size window is actually here ten days. 00:51.150 --> 00:54.870 So this is actually the rolling average. 00:55.140 --> 01:03.070 And actually in contrast to that with expanding windows our windows increase as we move forward in time. 01:03.120 --> 01:07.540 So at the very first time stem we have only one time for our calculations. 01:07.680 --> 01:11.410 Then at the second timestamp we have two and so on. 01:11.610 --> 01:19.500 And coding wise pandas actually provides us with the surprise expanding method so we can you see the 01:19.500 --> 01:26.610 expanding method on our clothes column and we can calculate actually the expanding mean. 01:26.610 --> 01:34.360 So let's say this here so here we are for the very first time stamp at the thirty first of December 01:34.360 --> 01:36.170 2000 and eight. 01:36.250 --> 01:39.760 And here we have the value of nine hundred three point two five. 01:39.970 --> 01:40.500 And uh. 01:40.540 --> 01:40.790 Yeah. 01:40.800 --> 01:47.920 This is uh simply the very first time stamp here in our close column and then at the second time stamp. 01:47.920 --> 01:51.430 So it's uh the second of January 2009. 01:51.430 --> 01:57.940 If you calculated the expanding mean then pandas simply takes uh the average of those two time stamps 01:58.510 --> 02:04.850 and then at the fifth of January 2009 pandas calculate the mean of um the. 02:04.870 --> 02:13.400 Immediately three proceeding uh timestamps here and so on so let's go down here now and that's actually 02:13.400 --> 02:14.870 for the expanding method. 02:14.880 --> 02:16.100 Uh one parameter. 02:16.100 --> 02:18.100 So let's have a look. 02:18.110 --> 02:20.890 So here we have the min period parameter. 02:21.440 --> 02:26.880 And if we can define the minimum number of observations in a window required to have a value. 02:27.020 --> 02:32.540 So the default setting is 1 but we can also set this for example to 5 02:39.640 --> 02:45.230 so we can see that we get the very first value at the fifth time stamp. 02:45.250 --> 02:47.320 So here we are in total five. 02:47.470 --> 02:50.330 Immediately preceding timestamps actually. 02:50.950 --> 02:54.250 But let's keep here the min periods per meter 2 1 02:58.510 --> 03:02.670 and 2 comparison Bill moving averages and expanding windows. 03:02.770 --> 03:05.860 We can create the two new columns. 03:05.860 --> 03:08.430 So the a simple moving average is 50. 03:08.950 --> 03:11.680 And uh the expanding window. 03:11.680 --> 03:16.970 So let's do this here and let's have a look. 03:16.990 --> 03:19.580 So uh here follow a simple moving averages. 03:19.690 --> 03:23.660 The first the 50 values are obviously any and values. 03:23.660 --> 03:25.280 So no surprise here. 03:25.510 --> 03:31.870 And now we can also visualize those two columns by using the other plot method. 03:32.350 --> 03:39.520 So here we can see in blue the simple moving average 50 and in green the expanding window. 03:39.520 --> 03:47.260 So at each single point in time here we have uh the average price for all of the preceding timestamps. 03:47.350 --> 03:51.490 And we can see here this is a quite stable line. 03:51.490 --> 03:59.250 And the same method with the rolling method you can do more than just the calculating expanding averages. 03:59.680 --> 04:08.020 So for example we can also calculate the expanding minimum stock price for example by changing here 04:08.050 --> 04:09.820 the main method so let's have a look 04:13.730 --> 04:20.690 so this this year the expanding minimum and obviously the total minimum here in the period between 2009 04:20.720 --> 04:25.940 and 2018 is here directly in the year 2009. 04:25.940 --> 04:32.560 And we could also visualize the expanding maximum with the max method 04:40.050 --> 04:43.090 so here we have in green the expanding maximum. 04:43.230 --> 04:53.850 And as our S&P 500 index this year increasing from 2009 to 2018 our expanding maximum see are also increasing. 04:53.850 --> 04:58.110 So these were fewer examples and applications of expanding windows. 04:58.230 --> 05:03.600 And there's actually a lot of more to discover here but for the time being we are finishing with this 05:03.600 --> 05:06.570 video and I hope to see you also in the next one by.