How would you build EUR Eonia, 6M Euribor and 3M Euribor curves without a global fitter? The main issue is that in the EUR market you have a 3s6s basis across futures and the actually liquid 6m Euribor swap market. Assuming the Eonia is built indepdently. How can you build a 6M Euribor curve based on (3s6s adjusted) 3M Euribor futures, 6M Euribor swaps? How can you then build a 3M Euribor curve by doing a 3s6s adjustment or however else, so that you end up with a curve that reprices 3M Euribor futures?Thank you or any help.Alternatively how would you build that curve sequence?m

Why not build your 6m curve with 6m FRAs and 6m swaps and your 3m curve - with euribor futures and 3m swaps? Et voila...

Maybe I am overthinking it, but where do you get 6m FRA and 3m swaps quotes from? If so, are they any good? I guess I am believing that only futures and 6m swaps, and 3s6s are a good starting point. Also, for risk considerations it's nice to drive everything in terms of futures and 6m swaps and express everything else as a basis to that.

They're good enough, in my experience... Frankly, I don't think that you're going to get a significantly better fit by screwing around with the basis quotes. Just my personal preference and I have played with multiple variations in the past (a long time ago, to be fair).I see your point about risk... I'd say, based on my perception of liquidity/accuracy of quotes, 3m curve (futures and 3m swaps) would be a natural base curve. I've always really hated messing with the basis in the short end.

I appreciate your advice and am inclined to believe you given your experience. Would you advocate the same for GBP?Actually I did a bit more digging yesterday and implying a 6M Euribor curve from a 3M Euribor curve comes at a loss of information. It's an averaging operation, hence reimplying a 3M curve from a 6M isn't going to look good. I probably need to define an IMM stub basis to ensure that the final 3M curve reprices futures. Not sure if the software I am using allows that...Or follow your route or build a global fitter myself...

TBH, depending on the sort of accuracy you require, the returns from this exercise quickly start to diminish... In general, I always prefer to err in favor of simplicity at the expense of accuracy (within reason, of course).

I agree with you in general, but that still depends on the client's preference. I have limited experience of using curves as a client, but I have some. Sometimes there is a need for known accuracy. In this instance, to be specific, 6M EUR FRAs are not liquidly traded instruments, so whatever you pick up, some dude has put this through a curve before sending it to Reuters or whoever. That could be sufficient still. Then there is the problem of how synchronous these quotes are. Anyway, I considered doing what you say, but the concern was with 6M FRAs mostly. I have no idea about 3m Euribor swaps, but I'd assume that the long 3s6s basis is less dynamic, so that disregarding that 3m Euribor swaps went through the dude's curve, 3m Euribor and 6m Euribor are awash. But maybe 6m FRAs are just fine as you say. I have no idea and should probably just do as you say and refine if needed.

QuoteOriginally posted by: mtsm[br]How would you build EUR Eonia, 6M Euribor and 3M Euribor curves without a global fitter? Why do you want to avoid global fitter?With multi-curve framework you are restricted yourself severely if you don't have a global fitter. As discuss below, the EUR example is probably the one with the most (reliable) data for 3M and 6M. The impact is less in that currency. But if you move to AUD (standard swaps are Fixed3M v BBSW3M up to 3Y and Fixed6M v BBSW6M from 4Y on), there is almost no way to avoid it. If you move to USD, you need also 2 curves simultaneous fitting if you use fed fund swaps. If you use cross-currency swaps (in a foreign currency collateral framework), you need to build 2 (or 3) curves simultaneously, usually from cross-currency swaps and fixed v 3M swaps (even it this is not theoretically perfect). Multi-curve (any number of them) simultaneous calibration are certainly the standard nowadays. And as you said from a risk management/hedging perspective, you want to calibrate your curves with the instruments you plan to use in practice.

errh, because I don't have one handy... and I am not going to write one just now... if I had one, I would use it.Other than that, while I agree with what you say, it is also a somewhat quanty way of looking at the world. The AUD case is an almost silly textbook case to me. I regret that so many people use its peculiarities as an argument to justify the existence and need for global fitter. I know that you have other good arguments for why it is good to use one, I am just picking up on AUD...That being said, one reason you would want to avoid a global fitter, is that you might want to fool around with market microstructure effects or I don't know what other experiments. It is possible to expose a truly complex models to excel or a scripting environment, but it's rather rare and complicated to do. Hence most global fitters are very much what you see is what you get and often not very accessible to clients.