RSI and others on tick data?

[dtc1969]

Hi,

I can calculate an RSI using closing prices etc. for end of day, but how do I do it using 1 minute, 5 minute tick data etc.?

For a 14 day RSI, do I need to add all upward / downward movements in closing price for EVERY tick in those fourteen days?

Or do I use matching times i.e. add all upward / downward movements in closing price at 12:30 from one day with 12:30 from another to get an RSI for 12:30?

This obviously applies to other indicators too.

Thanks for any help.

Darren

 

[dcraig1]

Its up to you really, but I would suggest that you need to construct OHLCV bars with say 100 ticks (or whatever) for each bar. Then timestamp the bar with time of the first tick in the bar. You end up with a time series, but it has bars of varying duration.

You can do exactly the same thing for constant volume bars (eg 100 contracts wide) but the V bit of the OHLCV ceases to have any meaning.

Your indicator code should then work on any sort of time series of OHLCV bars.

I have done this sort of thing in Java by extending a base OHLCV time series class for constant volume and and constant tick bars. I could extend it for constant range bars if I had any inclination to do so (Which I don't).

Putting some thought into your time series representation can save a lot of work later.

 

[dtc1969]

dcraig1,

Thanks for the prompt reply!

I assume that you're talking about real-time ticks as in:

17/05/2002:0829,5258.00,5258.00,5258.00,5258.00
17/05/2002:0829,5258.00,5258.00,5258.00,5258.00
17/05/2002:0830,5258.00,5258.00,5258.00,5258.00
17/05/2002:0830,5258.00,5258.00,5258.00,5258.00
17/05/2002:0830,5257.70,5257.70,5257.70,5257.70
17/05/2002:0830,5257.70,5257.70,5257.70,5257.70
17/05/2002:0831,5257.70,5257.70,5257.70,5257.70
17/05/2002:0831,5257.80,5257.80,5257.80,5257.80
17/05/2002:0831,5258.60,5258.60,5258.60,5258.60
17/05/2002:0832,5258.50,5258.50,5258.50,5258.50
17/05/2002:0832,5258.30,5258.30,5258.30,5258.30
17/05/2002:0832,5258.20,5258.20,5258.20,5258.20
17/05/2002:0832,5258.20,5258.20,5258.20,5258.20
17/05/2002:0833,5258.10,5258.10,5258.10,5258.10
17/05/2002:0833,5257.80,5257.80,5257.80,5257.80
17/05/2002:0833,5259.00,5259.00,5259.00,5259.00
17/05/2002:0833,5259.00,5259.00,5259.00,5259.00
17/05/2002:0834,5258.20,5258.20,5258.20,5258.20
17/05/2002:0834,5258.20,5258.20,5258.20,5258.20
17/05/2002:0834,5258.20,5258.20,5258.20,5258.20

where there's no regularity? And that would impose a pattern to get some meaningful data?

I have this and 1 minute, 5 minute, up to 60 minute period data.

My thinking is that for the FTSE I have 9.5 hours of data per day, and using 5 minute data, that is 12 prices per hour, 114 per day.

I'm using a thing called TA-SDK which has functions for all the indicators. However, if you specify a period of 14 and used the 5 min data, it would calc the RSI of every 14 lots of 5 mins! Not really what I want.

So, I'm thinking that if I use 14 x 114 (1596), that is the only way I can get an RSI for 14 days, because the total number of OHLCs for that period is 1596?

What do you reckon?

Thanks,

Darren

 

[uburs]

dcraig1,

Thanks for the prompt reply!

I assume that you're talking about real-time ticks as in:

17/05/2002:0829,5258.00,5258.00,5258.00,5258.00
17/05/2002:0829,5258.00,5258.00,5258.00,5258.00
17/05/2002:0830,5258.00,5258.00,5258.00,5258.00
17/05/2002:0830,5258.00,5258.00,5258.00,5258.00
17/05/2002:0830,5257.70,5257.70,5257.70,5257.70
17/05/2002:0830,5257.70,5257.70,5257.70,5257.70
17/05/2002:0831,5257.70,5257.70,5257.70,5257.70
17/05/2002:0831,5257.80,5257.80,5257.80,5257.80
17/05/2002:0831,5258.60,5258.60,5258.60,5258.60
17/05/2002:0832,5258.50,5258.50,5258.50,5258.50
17/05/2002:0832,5258.30,5258.30,5258.30,5258.30
17/05/2002:0832,5258.20,5258.20,5258.20,5258.20
17/05/2002:0832,5258.20,5258.20,5258.20,5258.20
17/05/2002:0833,5258.10,5258.10,5258.10,5258.10
17/05/2002:0833,5257.80,5257.80,5257.80,5257.80
17/05/2002:0833,5259.00,5259.00,5259.00,5259.00
17/05/2002:0833,5259.00,5259.00,5259.00,5259.00
17/05/2002:0834,5258.20,5258.20,5258.20,5258.20
17/05/2002:0834,5258.20,5258.20,5258.20,5258.20
17/05/2002:0834,5258.20,5258.20,5258.20,5258.20

where there's no regularity? And that would impose a pattern to get some meaningful data?

I have this and 1 minute, 5 minute, up to 60 minute period data.

My thinking is that for the FTSE I have 9.5 hours of data per day, and using 5 minute data, that is 12 prices per hour, 114 per day.

I'm using a thing called TA-SDK which has functions for all the indicators. However, if you specify a period of 14 and used the 5 min data, it would calc the RSI of every 14 lots of 5 mins! Not really what I want.

So, I'm thinking that if I use 14 x 114 (1596), that is the only way I can get an RSI for 14 days, because the total number of OHLCs for that period is 1596?

What do you reckon?

Thanks,

Darren

Hi Darren,

Can You please let me know where can i get this per tick ftse 100 data?

Thanks.