stochrsi returns incorrect values

[cryptorg-io]

Hi guys

I Have a problem with calculating Stochastic RSI.
I think this function Trader::stochrsi returns incorrect values. I've compared it on tradingview.

Thanks

[jb-lopez]

Do you have all of the parameters the same?
What are the parameters you are using?

[sixcom]

I got the same issue, according to TV, last output should be 24.8917(K), 29.4686(D) https://ibb.co/RpVC0qt

but it gives these

$close = [4265.36,4283.8,3774.99,3936.69,3731.32,3775,4225.03,4248,3976,4142.01,4103.19,3833.47,3901.84,3694.39,3433.26,3380.01,3401,3531.18,3410.15,3349.36,3430.24,3265,3195.71,3183,3195];

print_r(\LupeCode\phpTraderNative\Trader::stochrsi($close, 14, 3, 3));

Array
(
    [FastK] => Array
        (
            [18] => 44.961152562973
            [19] => 0
            [20] => 100
            [21] => 0
            [22] => 0
            [23] => 0
            [24] => 100
        )

    [FastD] => Array
        (
            [18] => 75.171844086438
            [19] => 48.320384187658
            [20] => 48.320384187658
            [21] => 33.333333333333
            [22] => 33.333333333333
            [23] => 4.736951571734E-15
            [24] => 33.333333333333
        )

)

[jb-lopez]

Thank you @sixcom that information does help.

The purpose of this project was to be able to provide the exact functionality of the PECL Trader extension for those that could not install it. And in that it is a success. Even when PECL Trader would provide the wrong data this project reproduces that exactly.

To return the correct data is ideal; however this project would become divergent from PECL Trader. It would no longer be a drop-in replacement.

I'm not against this, quite the opposite. It would allow me to clean up the code a bit. This project is a direct port form the C/Java version of ta-lib, and as ta-lib had one source for both C and Java and was written to be compatible with both compilers, it is not the best.

I just need to decide how to differentiate the divergences.

• Should I have the master branch match PECL Trader and another branch produce correct results?
• Should I have the master branch produce correct results and another Branch match PECL Trader?
• Should I not have a master branch and have the two branches for PECL and correct?
• Should I have two versions of the functions in the same branch, e.g. Trader::stochrsi and Trader::stochrsiPECL?
• Something else?

Whatever I decide, I won't be able to provide a fix before the end of the year. I might not even decide which direction is best in the next two weeks.

I'm open to input.

[sixcom]

@jb-lopez

Did not expect I could get a reply so quickly, thanks for that :)

I totally agree that you should keep this lib compatible with PECL extension, based on my research, Python community got a similar issue with TA-Lib, and someone found Tulip Indicators got the right results.

I think it might benefit the lib users if you can add a fix or something like LupeTrader extends Trader and overwrite stochrsi(), then Trader is the drop-in replacement and LupeTrader has the correction

[jb-lopez]

That is a great idea. I forgot that I could just extend the class.

Thanks @sixcom !

[jb-lopez]

@cryptorg-io @sixcom

I've added the LupeTrader::slowstochrsi function into branch release/1.2.0. The mapping to TV is

• $rsi_period: RSI Length
• $fastK_Period: Stoch Length
• $slowK_Period: K
• $slowD_Period: D

In my limited testing this appears to match up to the TV results nicely. I'll be doing some more testing later. If you wish to try it out and let me know how it works for you I would welcome the feedback.