FormulaBoinc 2017, Threats/Opportunities

ChristianVirtual

ChristianVirtual

[H]ard DCOTM x3
Joined
Feb 23, 2013
Messages
2,561
Following the FB statistics for the marathon are a bit unhandy plus not provide enough information about threats and opportunities.

To make our life a bit easier I wrote a little database importing the statistics from FB and use linear progression to calculate when we pass any trajectory of our competitors.

Right now the data points are rather few but a first shot and sample check showed some "real" threats.

Will try to make a nicer visualization next time around.

2017/8/14

Threats
BURP
[H]ard|OCP 2518
Vancouver, Washington 18 2017-08-14
Universe Examiners 6 2017-08-19
Brony@Home 8 2017-08-20
Gay USA 2 2017-08-29
BOINC@Austria 4 2017-09-13
BOINC@Heidelberg 0 2017-09-29 ​

Citizen Science Grid
[H]ard|OCP 25
christians 18 2017-09-03 ​

Climate Prediction
[H]ard|OCP 12
BOINC@Finland 0 2017-08-19
BOINC@Heidelberg 8 2017-09-03
BOINC@Austria 0 2017-09-27
South Africa 6 2017-10-05
Team niconico 1 2017-10-08
Gay USA 0 2017-12-08 ​

Enigma@home
[H]ard|OCP 25
BOINC@Heidelberg 18 2017-09-10 ​

GPUGrid
[H]ard|OCP 25
Microsoft 18 2017-11-11 ​

Leiden Classical
[H]ard|OCP 12
U.S. Army 2 2017-12-19 ​

LHC@home 1.0
[H]ard|OCP 18
christians 12 2017-11-15 ​

Milkyway@home
[H]ard|OCP 12
Brony@Home 10 2017-08-26
Richard Dawkins Foundation for Reason and Science 6 2017-10-08
BOINC@Denmark 8 2017-10-29 ​

Moo! Wrapper
[H]ard|OCP 15
Catalunya 4 2017-09-25
BOINC@Heidelberg 12 2017-12-07
Brony@Home 10 2017-12-19 ​

NFS@home
[H]ard|OCP 25
christians 18 2017-11-17 ​

PrimeGrid
[H]ard|OCP 18
Microsoft 15 2017-09-04 ​

Rosetta@home
[H]ard|OCP 12
U.S. Army 8 2017-08-20
christians 10 2017-08-21
Microsoft 6 2017-10-30 ​

Wanless Mersenne +2
[H]ard|OCP 25
christians 18 2017-09-03 ​

World Community Grid
[H]ard|OCP 25
christians 18 2017-08-15 ​



Opportunities

Seti@home
[H]ard|OCP 4
[email protected] 6 2017-10-16
​
Yoyo@home
[H]ard|OCP 10
BOINC@Heidelberg 15 2017-08-29
Polish National Team 18 2017-09-10
South Perth Hospital 25 2017-09-24 ​
 
Last edited:
Nice. Anyway you can set this up to show the stats in real-time on a web page?

I spend a good bit of time going through each project and comparing it to the daily averages for the teams in our league to see if we're above their averages or below them. If this data could be displayed in real time within' a web page that could be awesome.
 
Holy crap that's complicated looking! I think a table for each project would be better. Show us in the middle with 5 slots on either side showing threats and overtakes (unless we're already in first).
 
For WCG, Christians are doing ~400K PPD, while we are doing ~200K PPD. That's because most of us are doing some other projects.

The graph below kind of summarizes where we are based on last 60 days of crunching. Not good considering that WCG is one of our pet project.

Need some strategy thinking here. Fastgeek, need help here!

For SETI, we can catch up with [email protected], Gay USA and Bulgaria if we maintain out current output for quite some time for a measly 6 points gain at most.

Not trying to be negative here, but trying to be realistic....


upload_2017-8-15_21-55-11.png
 
Looks much better and I can actually understand it. Are you open to some suggestions on changes and things to add?
 
anytime ... just updated a colored agenda and specified what the color means.

Other suggestions are more then welcome
 
Okay cool.

You have 5 coloumns on each side, but it appears that only 3 are utilized in the overtake column? Unless it's not showing teams who are producing more than we are on average already maybe? For example, Seti@Home. We are in 8th place, to their should be 5 teams in front of us in that column correct? Unless it's showing ONLY teams we will overtake. If it is, let's change that. Can we somehow get it to show the averages in the boxes so we know how much more we need to produce to catch those in the lead?

The credit totals are displaying the numbers with ' instead of ,
For example 29'093'210 should be 29,093,210. Unless that's only an American thing?

Are you going to put all projects on there? Seems to be missing a lot.
 
CV, great table and good overview. For WCG, I didn't do the linear regression but at our current and top team RAC, we wouldn't be able to catch the top team with current trajectory, right? I'm just eye balling the graph that I showed using BOINCstats. So instead of "we might overtake in the future", perhaps we should state "need battle strategy to overtake in the future" or something similar.

With the table, we should focus on strategy for the next 3 or 4 months.

Just my one cent.
 
And if memory serves me right, FB in the past ended at the end of November. So that means, no xmas challenge. They may have changed that this year, but I don't really care....lol. So, we would only really have the team rallying for the WCG birthday challenge in November to help boost our standing. We typically can dig up some big hitters for a single challenge or two during cold months, but it would be counter productive to beg them during warm months and then beg them again two more times over the holidays...
 
Added a total view including all projects even if we currently have no chance to overtake.
Still to add the average credit as indicator for performance (needs).


And no, will not implement an auto pilot where the script runs and controls all our clients to optimize the outcome though that would be fun. :cool:
 
I can't tell much of a difference between the different colors. Might just be me though.
 
ok; let have a look on the newest version:
  • changed number format
  • got rid of the centered column and show all from point 25 to point 0
  • added average points (but still not enough data for good precision; need few more days data collection)
  • also include those teams in projects we have right now no chance
  • Added inactive indicator (zero gain for all teams)
Let me know if it create more clarity for you ...

Same link: http://fbradar.christianvirtual.online/
 
Last edited:
Wow looks good. Can we get some more darker colors for the color coding? It's difficult for me to see the different shades of colors.
 
Hmm, let me see. Those colors are predefined by the framework used underneath. But I'm sure there is some CSS file to adjust ...
 
CV gives us a good view of the threats and opportunities. Now we need to come up with some sort of a concrete battle strategy.

Here is my personal strategy/goal to support our team to beat Gridcoin in total FB points.

Starting with the opportunities first where we are well behind to gain top spot, I'm currently focusing on Yoyo (4th-12pts) and trying the The dDM-Project (5th-10pts). Assuming I'm the only major contributors to these two projects, I'll be pretty happy if by the end of Nov (?), we could get gain 13 + 15 = 28 pts for the team. So now, I've stopped running WOW (hot + electricity cost) and focus on these. I think I have a good chance in Yoyo but will try my luck in The dDM-Project. Will see.

On the threat side, I'll look into this during the cooler period and see where I stand on these two projects.


upload_2017-8-17_21-41-25.png


upload_2017-8-17_21-41-55.png
 
CV, just to clarify.

What is g-bar? Is that average daily gain?
Underneath this you have a date. I suppose this is the date at the average current run rate in which we will overtake the team and the number in brackets is the number of days to overtake? Correct?
 
pututu said:
What is g-bar? Is that average daily gain?
Click to expand...
correct
pututu said:
Underneath this you have a date. I suppose this is the date at the average current run rate in which we will overtake the team and the number in brackets is the number of days to overtake? Correct?
Click to expand...
correct; with the current run rate based on the "best-fit-line"/linear regression and determination when "our" line cut "their" line.
 
pututu said:
I'm currently focusing on Yoyo (4th-12pts) and trying the The dDM-Project (5th-10pts).
Click to expand...

I have also right now a EC2-spot instance on yoyo / CSG with 36 cores. Not sure how long I will run it though.

My two Ryzen with total 32 cores are on WCG; it just hurts that we lost the place. Not that I can make a bit jump into it.

GPUs remain on WOW; no risk on GPUGrid right now.
 
ok; changed my mind. in WCG we can't loose too quick another spot; but in Rosetta@home. So moving there with my Ryzen to try to keep the follower on distance.
 
Keep in mind that the WCG birthday challenge is in November. It is bigger than the Wow Event. The team typically participates in that and then in the Xmas race through WCG's challenges throughout December.
 
ChristianVirtual said:
I have also right now a EC2-spot instance on yoyo
Click to expand...
Awesome man.

I'm splitting my rig on Yoyo and dDM-project but will try to gain more points in dDM-project.

At least I can see some green light at the end of the tunnel this morning for dDM. Did you manage to find the part where you need to change the CSS color codes? This is not urgent as my eyes are still ok :)

upload_2017-8-18_7-9-3.png


I think we should attack those few low hanging fruits first as it gets harder to displace top teams in other projects (e.g. Collatz, SETI, Rosetta, etc) . Hence we may need to act like a wolf-pack or lay out alternative strategy here or just keep crunching whatever you are doing now. I would prefer team consensus....
 
Skillz are you still working on Primaboinca? Doesn't seem like there is much of a gap but that may be because of project issues. On the other hand, may be easy points. I personally don't think they should be part of FB but that is just me.

Looks like we are 300k away from 15 points and last I looked you were like 30k/day. If we took first there, we would still need 37 points to tie and 38 to lead.
 
Last edited:
pututu said:
Did you manage to find the part where you need to change the CSS color codes? This is not urgent as my eyes are still ok :)

View attachment 33756

I think we should attack those few low hanging fruits first as it gets harder to displace top teams in other projects (e.g. Collatz, SETI, Rosetta, etc) . Hence we may need to act like a wolf-pack or lay out alternative strategy here or just keep crunching whatever you are doing now. I would prefer team consensus....
Click to expand...

not yet spend too much time finding the CSS file for it; also I would need to change and not use the CDN version or overwrite the definition. Anyway; still on the todo list.

As for the strategy: team consensus would make more sense.

Primaboinca are really low hanging fruits; I tried to sign up but no chance (if once with an account could try to contact the admin to get my account setup ?)

Rosetta should get some attention to defend our position and maybe even catch up with the BOINC-waffle team.

Enigma will be a big lost if we don't keep the spot


Do we have a math-guru could help to define a model to automatically suggest the best allocation (would also require the tracking of capabilities of individuals); sounds something like "Operations Research" to me (I'm too long out of school to remember; but I loved it)
 
Yeah I am working on PRIMABOINCA. Project was down for ~3 weeks but it's back so I'm trying to get what I can from it. Too bad it only runs on Windows though. I think I might have a spare HDD laying around that I can get another rig running it as well. (10c20t Intel) that should double my production on it.
 
ChristianVirtual try vndi83nodi3E That is what Skillz recently got in with. It may change so report back your results please.

If you email them from the email on the home page anyways please do. I'm interested in how long it takes them to reply if they do at all.

I'm still running ODLK, duchamp (when they get work again) and drugdiscovery primarily until I hit the 1 million point goals. ODLK is in my sites for testing for DC-Vault.
 
The invitation code worked , got my account created manually. Now the info that it Windows only is setting me back as I'm Linux only (for any DC activity; no plan to torture my only windows laptop or Mac with BOINC or FAH)
 
Yeah I feel your pain. All my 4P rigs are Linux only so I'm having to run this on my single CPU Windows boxes. One of those was even running Linux, but I found a spare HDD so I am dual booting it Linux/Windows since I'm not using it for Linux at the moment.
 
I just signed up Primaboinca in case they stop accepting new donors.
 
Can run Windows from within a VM..lol...that is how I ran WEP-M+2. Linux in a VM on top of Windows.
 
I might try this with one of my 4Ps. Didn't think about using a VM within' Linux to run Windows 7.
 
I just wish the ReactOS guys would get their shit together and put out a usable product. Last I checked their OS only used like 90MB or so of RAM. However, BOINC didn't work on it back then. Gonna try their latest build hopefully in the upcoming week.
 
Hell yes! Now I can very, very easily just browse where we need boxen and where we don't. Good work man.
 
The green is a bit confusing though. For example. In PRIMABOINCA our average is well above the teams ahead of us, yet they're labeled green as if we need to come up with some strategy to overtake them. This isn't the case. We're going to in a few days so it's not a concern as long as the trend continues w/ our 60k+ PPD and their less than 5k PPD.

To me it should be teams that are ahead of us AND their current production is beyond ours.
 
The green is more to indicate: if we keep running the same performance we get them within this year.

If they are ahead of us with higher performance they are not colored but have a text label that we need to push ourselves harder.

Like in SETI, the green once are possible, more to the top we can't reach with our current performance. (And the green one also not as after WOW the attention will fall down again)
 
You must log in or register to reply here.
Back
Top