Rank: Member
Groups: Registered, Registered Users, Subscribers Joined: 3/15/2024(UTC) Posts: 15
Was thanked: 2 time(s) in 2 post(s)
|
Good morning, the results of the system test do not match those of the explorer.
Using the ACT series it is evident, the MAMA-FAMA system gives different results even though the formulas are the same.
So I wonder what are the "right" signals to follow?
Thanks
Valter
|
|
|
|
Rank: Advanced Member
Groups: Moderators, Registered, Registered Users, Subscribers Joined: 10/8/2010(UTC) Posts: 1,960
Thanks: 91 times Was thanked: 155 time(s) in 150 post(s)
|
Originally Posted by: temperino Good morning, the results of the system test do not match those of the explorer.
Using the ACT series it is evident, the MAMA-FAMA system gives different results even though the formulas are the same.
So I wonder what are the "right" signals to follow?
Thanks
Valter
Hello, The formulas are primarily the same although the Explorer has a few additional functions to limit "repeat" signals. For example, if you copy/paste the Long/Buy order from both System Test and Explorer, you will find they both spike to 1 (a true condition) at the same time, but the System Tester one can remain true for multiple days whereas the Explorer returns back to zero immediately after the spike. When you get into a trade in the System Tester is based on multiple factors. Within the "Trade Options" of the System Tester under the "Trade Execution" tab, "Realistic Market Prices" is typically enabled, which will delay entry into a trade by 1 bar. This is designed to simulate more realistic conditions where you do not get into a trade on the bar of the signal but get in on the bar after the signal. You can of course uncheck this and set the "Delay order opening" option to 0. It is also recommended to not use "Load Minimum Records" in the Explorer when using certain complex and/or exponential calculations. It is recommended to set the Load Records to the same amount that you are using in other tools in these situations. This is typically more relevant when you are comparing signals in a chart vs. the Explorer, but you might want to try loading the same records in the Explorer (i.e. 1250) to see if this changes the results. Load Minimum Records can be a useful feature when running simple calculations and you want to generate results more quickly.
|
|
|
|
Rank: Member
Groups: Registered, Registered Users, Subscribers Joined: 3/15/2024(UTC) Posts: 15
Was thanked: 2 time(s) in 2 post(s)
|
Hi tanks for the answer
all the conditions you described have been verified but the result does not change, let's go into detail.
let's talk about the ACT MAMA FAMA system..
I do a retroactive weekly exploration indicating June 7 2024 as the day to explore, therefore indicating the specific dates,
the buy and sell indicated are also visibly confirmed by the "Exploration instrument column data" function, I mark 1 on June 7 and 0 for the previous 10 weeks, therefore it is a unique and unmistakable signal.
So I open the chart directly and I find the days of purchase and sale indicated on the graph.
Now I use the same system, ACT MAMA FAMA, for a system test. I check that both the formulas present in both the explorer and the trading system are the same, excluding the display parameters.
I set the period to test from January 1 2024 to today, therefore abundantly including June 7.
Result? the purchase of June 7 disappeared and reappears a week later....
if the system test is "right" then the explorer is not usable
if the explorer is "right" then it is difficult to make truthful backtests.
Only the ACT series has this behavior or other series too?
Thanks
very worried user...
|
|
|
|
Rank: Advanced Member
Groups: Moderators, Registered, Registered Users, Subscribers Joined: 10/8/2010(UTC) Posts: 1,960
Thanks: 91 times Was thanked: 155 time(s) in 150 post(s)
|
Originally Posted by: temperino Hi tanks for the answer
all the conditions you described have been verified but the result does not change, let's go into detail.
let's talk about the ACT MAMA FAMA system..
I do a retroactive weekly exploration indicating June 7 2024 as the day to explore, therefore indicating the specific dates,
the buy and sell indicated are also visibly confirmed by the "Exploration instrument column data" function, I mark 1 on June 7 and 0 for the previous 10 weeks, therefore it is a unique and unmistakable signal.
So I open the chart directly and I find the days of purchase and sale indicated on the graph.
Now I use the same system, ACT MAMA FAMA, for a system test. I check that both the formulas present in both the explorer and the trading system are the same, excluding the display parameters.
I set the period to test from January 1 2024 to today, therefore abundantly including June 7.
Result? the purchase of June 7 disappeared and reappears a week later....
if the system test is "right" then the explorer is not usable
if the explorer is "right" then it is difficult to make truthful backtests.
Only the ACT series has this behavior or other series too?
Thanks
very worried user...
Hello again, Which instrument are you running the test on in particular? What are you using to verify the trade in the System Test (i.e. "Plot on Chart")? If you double-click on the instrument in the System Test results and click on the "Orders" tab, you can scroll through the list of potential trades to see what has occurred. For example: Code:41 6/7/2024 37 Considered Sell Short 30 Market Signal
41 6/7/2024 37 Placed Sell Short 30 Market Signal
42 6/14/2024 37 Opened Sell Short 30 Market Signal
42 6/14/2024 37 Cancelled - Position Limit Sell Short 30 Market Signal
Note that the trade was not executed (Cancelled - Position Limit), because the System Test was already in a trade. This is controlled on the "General" tab of each system test. In particular, when Editing a System, the General Tab has an option called "Position Limit" and by default this is enabled and set to Limit number of simultaneous positions to "1". You can of course uncheck this option or increase the numerical value.
This is not something that happens with the Chart Template or the Explorer. Additionally, note that due to "Realistic Market Prices" being enabled, the potential trade was delayed until the next bar (on weekly data this means the trade is delayed until the next week).
Edited by user Tuesday, August 27, 2024 2:12:52 PM(UTC)
| Reason: Not specified
|
|
|
|
Rank: Member
Groups: Registered, Registered Users, Subscribers Joined: 3/15/2024(UTC) Posts: 15
Was thanked: 2 time(s) in 2 post(s)
|
Hello,
is not a metter of "repeat" signals, or trade execution and realistic market prices, all this stuff is set correctly. It's not a question of signals shifted in time, the signals in system tester and explorer are different.
to verify this anomaly:
run system test ACT MAMAFAMA from 1/1/2023 on 1 security (random stock)
in the system test/result detail/position detail there are all the trades with the entry and exit dates.
Run explorer for the same security and enter an exchange date equal to one indicated in the system test, I should get a buy or sell signal on that date but it is not so
if I change the entry date with another calculated by the system test the result will always be the same, they do not match
I tried with the help desk but they limited themselves to linking me to the ACT help...
Is it possible to post pic from the computer?
Now I will wait for the Italian importer to return from vacation and I will finally be able to show this anomaly
P.S. is this perhaps the problem?
A trend following strategy based on the crossover of MAMA and FAMA. ** DISCLAIMER ** - This tool is for not guarenteed for accuracy or correctness of any kind, and is experimental.
|
|
|
|
Rank: Advanced Member
Groups: Moderators, Registered, Registered Users, Subscribers Joined: 10/8/2010(UTC) Posts: 1,960
Thanks: 91 times Was thanked: 155 time(s) in 150 post(s)
|
Originally Posted by: temperino Hello,
is not a metter of "repeat" signals, or trade execution and realistic market prices, all this stuff is set correctly. It's not a question of signals shifted in time, the signals in system tester and explorer are different.
to verify this anomaly:
run system test ACT MAMAFAMA from 1/1/2023 on 1 security (random stock)
in the system test/result detail/position detail there are all the trades with the entry and exit dates.
Run explorer for the same security and enter an exchange date equal to one indicated in the system test, I should get a buy or sell signal on that date but it is not so
if I change the entry date with another calculated by the system test the result will always be the same, they do not match
I tried with the help desk but they limited themselves to linking me to the ACT help...
Is it possible to post pic from the computer?
Now I will wait for the Italian importer to return from vacation and I will finally be able to show this anomaly
P.S. is this perhaps the problem?
A trend following strategy based on the crossover of MAMA and FAMA. ** DISCLAIMER ** - This tool is for not guarenteed for accuracy or correctness of any kind, and is experimental.
Hi again, The disclaimer doesn't have anything to do with how MetaStock handles formulas. Formula functions are just like math. The program is going to handle the functions the same in each tool, all else being equal. Of course, not all is equal because the different tools are not designed to operate the same way, since they load data differently and have other criteria for getting into trades. This is why it is helpful to get all the details (what interval is being used, what symbol is being used, how much data is being loaded, etc.).
For example, with Realistic Market Prices Unchecked and Delay Order Opening set to 0, I ran a system test with 1250 Records on Daily Data for IBM. I then used the "Plot on Chart" function from the System Tester, and attached the ACT - MAMA/FAMA Expert to the Chart. Expert Buy/Sell Arrows align with the System Tester Orders. There was a Sell Order on Aug 6th, 2024. I then set the Explorer to run on Aug 6th w/Load 1250 Records and Daily Data for IBM. The Exploration Report gives me a "Short" column value of 1.000 (which aligns with the expert/system test). There was a Long Entry the next day (Aug 7th) which is reflected in both the Expert signal and the System Test Order. I then set the Explorer to run the scan on Aug 7th and the Exploration report gives me a "Long" column value of 1.000 (which aligns with the expert/system test). I do see that you contacted our chat support. We can continue to work with you through chat if that is easier. I have also created a record that links to this forum post in case there is any confusion as to the problem. Edited by user Wednesday, August 28, 2024 3:38:53 PM(UTC)
| Reason: Not specified
|
|
|
|
Rank: Member
Groups: Registered, Registered Users, Subscribers Joined: 3/15/2024(UTC) Posts: 15
Was thanked: 2 time(s) in 2 post(s)
|
Thanks for the reply, so it seems to be a problem only for me. By the way this anomaly also occurs with the PS series, not all systems.
I tried with the help chat but I only got links to the help present in MS. I will wait for the Italian importer to arrive from vacation and then we will see together how to solve the problem. Thanks for now, greetings Valter
|
|
|
|
Rank: Member
Groups: Registered, Registered Users, Subscribers Joined: 3/15/2024(UTC) Posts: 15
Was thanked: 2 time(s) in 2 post(s)
|
"I do see that you contacted our chat support. We can continue to work with you through chat if that is easier"
an intervention directly on my pc would be very useful, can we do it?
|
|
|
|
Rank: Member
Groups: Registered, Registered Users, Subscribers Joined: 3/15/2024(UTC) Posts: 15
Was thanked: 2 time(s) in 2 post(s)
|
I'm connecting to the chat now
|
1 user thanked temperino for this useful post.
|
|
|
Rank: Advanced Member
Groups: Moderators, Registered, Registered Users, Subscribers Joined: 10/8/2010(UTC) Posts: 1,960
Thanks: 91 times Was thanked: 155 time(s) in 150 post(s)
|
Originally Posted by: temperino Thanks for the reply, so it seems to be a problem only for me. By the way this anomaly also occurs with the PS series, not all systems.
I tried with the help chat but I only got links to the help present in MS. I will wait for the Italian importer to arrive from vacation and then we will see together how to solve the problem. Thanks for now, greetings Valter
Hi again,
There may be an issue, but the issue might only manifest under certain conditions. It could be related to the amount of data history that is being loaded into each tool. For example, with certain Exponential calculations, this can change the results of a signal depending on how much data you load because Exponential calculations carry a piece of the history forward, which could change whether a signal is true or not.
But this is why it will be helpful in resolving this issue to get all the specific settings being used in each tool. If you do contact us in chat again, you might want to link to this forum post so the chat representative can see what we have discussed here.
|
|
|
|
Rank: Member
Groups: Registered, Registered Users, Subscribers Joined: 3/15/2024(UTC) Posts: 15
Was thanked: 2 time(s) in 2 post(s)
|
good morning, today with the help of the importer we managed to solve the problem. 1250 records were not enough to calculate the entry and exit points exactly, we set them to 32000.
Thanks
regards
Valter
Edited by user Tuesday, September 10, 2024 3:55:11 PM(UTC)
| Reason: Not specified
|
1 user thanked temperino for this useful post.
|
|
|
Users browsing this topic |
|
Forum Jump
You cannot post new topics in this forum.
You cannot reply to topics in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.