I think a PMC for everything should only be used for fatigue. Individual PMCs show fitness/form in that particular sport. A combined PMC wont show much in terms of fitness as it doesn't show what you are fit for, but the accumulated fatigue is relevant.

I made separate graphs on my Dashboard for Bike and Run as well as one for them combined.

I have never put swim metrics on a graph.

First, in the PMC there is no such thing as "fatigue" (or "fitness", or "form") - only TSB, CTL, and ATL.

Second, don't combine data from different sports.

While I won't argue with the authority on combining.

On the Training peaks website, if you create a PMC chart on the Dashboard, Fatigue (ATL), Fitness (CTL) and Form (TSB) are labels on the graph when the mouse pointer is near any of the lines.

Are there any papers about fatigue balance whilst multisport training ie overtraining. Whilst the various TSS, sTSS, rTSS, hTSS are non-matching units and can't be combined, I'd like a better understanding on the interactions even on a general level.

I doubt (but hope I'm wrong) you'll get much evidence based. It's easy to say 'the units are non comparable' and in an empirical way this I'd right. I'll not be as 'trained' for cycling with a bCTL of 80 compared with a rCTL + bCTL of 80 but I doubt many people would argue that rTSS won't impact of biking performance. For an N=1 my cycling performance has markedly increased in the last few weeks with a decrease in rATL due to a minor running stress injury. bTSS increasing to fill the gap and hence bATL increasing. I don't count swimming exertion as anything outside swimming as I'm crap and it's basically a rest day for the CV system when I flail around in the pool for an hour burning 500kCal.
TLDR: I doubt you'll get a meaningful conversion formula but keeping an eye on total load seems prudent if for nothing else but retrospective analysis of performance.

I understand that, and my guess is also that there's no decent research out there.

My issue, which is a personal one, is that I'm trying to separate mental fatigue from physical fatigue. So the more accurately I can determine my physical fatigue levels, the more I can tell my brain to shut up or listen to it as appropriate . I do, rightly or wrongly, use a combined TSS score in the abscence of anything better.

I completely agree here. This year hasn't been the best of starts for me running wise, but has been superb for the bike (7+ hours a week, every week, for 12+ weeks).

In spite of the fact that, when compared to last year, I have: (1) run significantly less; and (2) the running I have done has not been as focused, my running is better than it was this time last year. I can currently comfortably run in z2 at a pace that I only achieved in April/May last year after 4 months solid run training. Yet, my running CTL is lower than it was last year. My combined CTL is a lot higher due to the fact that my bike CTL is a lot higher than it was this time last year.

That's why I have a run PMC, bike PMC and combined PMC showing on my training peaks as I am absolutely convinced that fitness in one has an effect on fitness in the other, even if my PMC for that specific discipline doesn't show it.

And, as per the previous poster, I don't want to argue with the foremost expert on this matter, but why does training peaks list CTL, ATL and TSB as Fitness, Form and Fatigue if there is no such concept???

I can't control what TrainingPeaks does. I can, however, share my $0.02 in forums like this one.

Is your position that running/cycling load has no bearing on the performance of t'other or are you just saying you ought not to add one to the other?
Another example I use is that if I come away from a marathon focus block where my bike CTL is pants then I'm markedly fitter than say I took a month off everything to arrive at the same CTL. To my mind that proves that aerobic work is transferrable to some degree even if the specificity benefit hasn't been gleaned. I'm only really interested in an injury or burnout predictor. Not an objective conversion factor and combined score.

There's this (cited in my introductory article about the PMC, BTW):

https://www.ncbi.nlm.nih.gov/pubmed/11774068

Among other things, what this study demonstrates is that the time constants are different for running and cycling, even in the same individuals. Thus, even if you could confidently express the demands of each on the same scale, entering combined scores isn't going to work as intended.

The issues are even greater if 1 TSS != 1 rTSS (which, when you consider the differing demands between the two sports, and that the two metrics were developed by different people and are calculated differently, seems more likely than not).

My advice therefore remains the same as it has always been: 1) don't attempt to combine running and cycling into one PMC, and 2) if you're going to ignore that advice, use Foster's session RPE as the input function.

The latter. Obviously there is some overlap/interaction (both positive and negative) between training for the two sports.

The same reason that they will tell you that FTP = maximal 1.0000000000000000000.... power, or that CTL is the average of your TSS over the last 42 d.

Thanks Andrew - really interesting point. I did always wonder where the 42 day calculation derived from. I have read Joe Friel where he says "CTL is the average of your TSS over the last 42 days, or whatever other time period you choose" and wondered why 42 was such a magic number. It appears that it isn't.

Very interesting.

CTL is calculated using an exponentially-weighted moving average using a 42 d (by default...this is adjustable in WKO+ 3.0/WKO4; I don't know about TP online) time constant.

What that means is that it mostly (i.e., ~90%) reflects what you have done for the last ~ 3 mo.

(42 d time constant = 29 d half-life; 3 half-lives = 87 d = 87.5% of the way to plateau.)

Same logic applies to ATL, which mostly reflects what you done for the last ~2 wk.

I'd read that as swimming has essentially no impact so can be treated in isolation.

Furthermore, whilst there are interactions between running and cycling, the effect isn't particularly pronounced so adding small increases in volume/intensity to one isn't going to greatly affect the other.

Or in summary, I need to HTFU.

ok i won't combine bike and run in the same pmc. But i can't understand the sense of a separate PMC to undestand my training load. what i'd like to obtain from this kind of tables is to understand if i'm tired before my body tells it to me. manage training volume is the key for a triathlete but we need to consider all the workouts.

a run pmc could indicate me that the X day i can go out for an hard vo2max session because this pmc don't consider that the day before X i was out for a really hard bike session.

All the fun of definition games, personally I like to track when #FF00FF is higher than #00FFFF & #FFFF00 drops.

I honestly thought you'd used those terms before, but turns out I was remembering Friel: http://www.joefrielsblog.com/...-balanceso-what.html

You come close here, but just draw parallels: http://home.trainingpeaks.com/...-performance-manager