simulation2

TEST 2

Precision Stop (6-12) Multiple test on huge data set

These tests are on the new version of the simulation unit with correct profit feature.

Risk profile set to 0.25 of account in all tests
Gearing max in all tests = 4.5
Equity starting at £10,000 all tests
Slippage set to 20% (note in these initial test 0.4 multiple gave a huge profit way above the slow settings) slippage is too low i think.
Spread 1.5 per side
Max gearing( in settings) 7
Max trader day 123
Min percent 0.01
Max percent .20 this will affect slow system settings and may need increasing>>>>>>>>>> NOTE!

Testing of Precision stop (6-12) Para multiple on 123 various stock, I have chosen only a few good system stocks, the rest are chosen because they were known to me. Ten years data tested.

Test run from Jan 3rd 2000 to August 12th 2010

These tests used the same raw database and copied and pasted it and ran indicator builder over it each time.

Candover Investments, JKX oil and gas, Quintain estates, Pendragon, JJB sports, Johnson press, Royal bank of Scotland, Lloyds TSB, Unite group, Polar capital tech trust. and various others...see excel list of stocks used

Before the test, I predict 0.83 multiple will work best....lets see.

Purpose of this test is to ascertain the best GLOBAL multiple settings in a simulation run over realistic portfolio over a large amount of sample.

Test	Multiple	Max cost	Trade div	Profit	Percent	Max dd	Max UW	Win-L	R-R	Trades	Gearing	10 Year MAR	Slippage
1	0.8	4.5%	70	£4,685,918	46,685%	25.05%	230	.42	1.89	13,281	4.479	1836	20%
2	1.8	5.8%	44	£4,815,195	48,051%	32.37	352	.4244	2.13	2945	4.41	1484	20%
3	0.4	3.8%	83	£23, 410,332,854	267,450,459%	16.3	94	.416	2.32	35,643	4.50	unreal	20%

Basis of the above testing, with Multiple 0.4 making £100million( previous test1 old simulation version) and now making £23 billion, one can assume that slippage of 20% is way out so retesting at 30% ( This is an attempt to rectify the difference between the bug in the previous version and the 50% applied to those tests. After performing some comparisons on the various slippage tests below the decision has been made to use 50% on all new tests.



3a	0.4	3.8%	83	£23, 410,332,854	267,450,459%	16.30	94	.416	2.32	35,643	4.50	unreal	20%
3b	0.4	3.8%	83	£2,505,575,293	25,025,652%	17.89	102	.403	2.27	35,771	4.55	unreal	30%
3c	0.4	3.7%	84	£180,123,195	1,801,131%	18.59	179	.391	2.32	35,838	4.48	unreal	40%
3d	0.4	3.7%	84	£13,999,496	139,895%	22.71	205	.377	2.20	35,910	4.52	6160	50%
3e	0.4	3.7%	84	£1,148,834	11,388%	29.08	298	.363	2.15	35,973	4.56	1484	60%

Test	Multiple	Max cost	Trade div	Profit	Percent	Max dd	Max UW	Win-L	R-R	Trades	Gearing	10 MAR
Something is wrong as the 0.3 setting, which would normally be unusable is making £52 billion quid...

8	0.3	3.8%	95	£52,077,074,473	520,770,644%	16.68	100	.371	2.47	44,086	4.517	1484	50%
3d	0.4	3.7%	84	£13,999,496	139,895%	22.71	205	.377	2.20	35,910	4.52	6160	50%
4	0.5	4.05%	81	£768,070	7,580%	29.87	298	.384	1.93	26,985	4.581	1484	50%
5	0.6	4.3%	72	£163,855	1,538%	47.37	667	.394	1.75	20,381	4.812	1484	50%
6	0.7	4.4%	70	£78,876	688%	50.65	715	.396	1.70	16,076	4.696	1484	50%
7	0.8	4.3%	70	£60,251	502%	40.75	732	.402	1.65	13,305	4.41	1484	50%
9	0.9	4.5%	65	£74,766	647%	45.22	1360	.406	1.69	11,301	4.55	1484	50%
10	1.0	4.6%	63	£99,576	895%	40.9	937	.409	1.68	9,871	4.65	1484	50%
11	1.1	4.6%	63	no data yet	895%	40.9	937					1484	50%
12	1.2	4.7%	60	£110,573	1,005%	47.4	1281	.409	1.74	8095	4.65	1484	50%
13	1.3	4.7%	60	no data yet	1,%	47.4	1281	.409	1.74	8095	4.65	1484	50%
14	1.4	4.7%	59	£135,332	1,253%	57.12	834	.403	1.80	7109	4.53	1484	50%
15	1.5	4.7%	60	no data yet
16	1.6	4.7%	60	no data yet
17	1.7	4.7%	60	no data yet
18	1.8	5.6%	44	£600,000+	5,889%	45.53	620	.4104	1.95	2945	4.44	1484	50%
19	1.7	4.7%	60	no data yet
20	2.0	5.6%	44	£138.999	bug trades	83	max per?	.4104	1.95	5744	4.44	1484	50%
21	2.0	5.0%	45	£6,014,735	bug trades	41	max per75%	.409	2.07	2341	3.91	1484	50%
22	2.5	5.0%	5	£441,642	bug trades	35.98	max per50%	.410	2.08	1819	3.33	1484	50%

Some ideas of this error...

Could it be that the faster system is getting tripped up by giving 2 orders in one day?

BUG TO NOTE. RAN DATA COLLECTION BUT NOT INDICATOR BUILDER , OPENED SIMULATION TO TEST DONCHIAN, BUT IT SAID NO DATA TO SIMULATE.

2nd time loaded up some 0.4 data and the donchian crashed after 2720 days with mess no row at position 0.

I will try again from scratch...

2. system is adding trades on day 1 instead of waiting for stance change.

3. busy system tripping up

4. Percentage tests see bug 5 ( this also occurs in large multiple settings)

5 big bug when running 50% system, symbol CCL computed a stop of 504974.3 from an entry at 861 and a bid of 1347.64 and as we are now

using MAX to calculate the profit that gets booked, it booked £665,429 profit ( incorrectly on that trade ) date 01/07/2002

6, day counter was at 75 or 76 then went down to around 5, then back up to 77 again, (same data bed as before around 16th 04-2000)

later it jumped by several hundred points without explanation. for many dates after it is jumping all over the place, and so is the days tested window.

7. the system is putting on all longs at the start ( which is probably what i said we should do initially in 2005) this time i want it to wait for a stance change before doing anything.

Expectations from £10,000 over ten years...

Per anum growth rate	Result
5%	£16k
10%	£26k
15%	£40k
20%	£62k
25%	£93k
30%	£138k
35%	£201k
40%	£289k
45%	£411k
50%	£576k
55%	£800k
60%	£1.1 million
65%	£1.5 million
70%	£2 million
75%	£2.7 million
80%	£3.5 million
85%	£4.7 million
90%	£6.1 million
95%	£7.9 million
100%	£10 million+
200%	£590 million+

Initial conclusions.

Target gearing has been around 4.5 times equity, but the above tests are not exact. During these tests the highest gearing only shows during the last few years of the test when the full data set is active.

Settings used are revealing a surprisingly high frequency system to be working best. The overall impression shows best values lie between 0.7 and 0. 4 with the likely best being between 0.45 and 0.55.

Trading with this set up the system generates 20,000 to 30,000 trades over a 10 year period which on 123 stocks equates to around 2,500 trades per years / 250 = 10 trades per day.

This seems a higher frequency than I originally expected.

Paradoxically the best profit settings between 0.4 and 0.6 gave very poor underwater readings.

What to do next? Mull over the results and go in the direction which can produce a model that I can stick to. Enduring more than 200 days underwater is not something I wish to do and neither do I want to have more than 20% drawdowns.

Basis of this 0.6 to 0.7 seem to look favourable. Next step is to test in 0.025 increments over the best area and see what comes up.

It would also be wise to test 1.25, 1.5, 2.0 and 2.5 just to be sure there isn't a double top shaped distribution curve hiding on the chart below.

Hence, now testing 1.2, 1.4, 1.6 etc if no improvements after that will abort long term testing. I am not expecting great results from the slow model.

last test was 81 divisor and 4.05% max trade, and had 4.53 gearing, so its still too high

Vertical axis = Millions x 10 New test 2 Horizontal axis = Multiple setting
520


21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
1
	1.8	1.6	1.4	1.2	1.0	0.9	0.8	0.7	0.6	0.5	0.4	0.3

Vertical axis = Millions x 10 Old test 2 Horizontal axis = Multiple setting
13
12
11
10
9
8
7
6
5
4
3
2
1
1
	1.8	1.6	1.4	1.2	1.0	0.9	0.8	0.7	0.6	0.5	0.4	0.3

Vertical axis = Millions x 10 Horizontal axis = Multiple setting
13
12
11
10
9
8
7
6
5
4
3
2
1
1
	1.8	1.6	1.4	1.2	1.0	0.9	0.8	0.7	0.6	0.5	0.4	0.3

Vertical axis = Millions x 10 FINE TUNING TEST (New tests in green) Horizontal axis = Multiple setting
18
17
16
15
14
13
12
11
11
10
10
9
8
7
6
5
4
3
2
1
1
	.34	.36	.38	.40	.42	.44	.46	.48	.50	.52	.54	.55	.56	.58	.60	.62	.64